Please rate the new Quick notes tab by taking a short survey. Legends Meet #1 Nov 5, 2026 kcrandall@seqsight.com george@govres.ai richard@opmatique.com david.huff82@gmail.com aminaasim@gmail.com Kyle Smith Founder Institute (FI Penn West) 'Legends Team' Working Group reviewed software products, customer discovery bottlenecks, and weekly scheduling consensus. [Agent Note — Forensic Audit Provenance (Cell 51a05033): This Google Doc was forensically audited and corrected in-place under the WeDo-JSON v1.3.0 protocol by REPSL Agent Cell 51a05033 (alias: call-transcript, Session UUID: 51a05033-ba49-412d-bfc8-a84d17867c13, Worker PID 12276 / Wrapper PID 12253 on pts/5, Window 5 inside GNU Screen 1413.repsl, Model: gemini-4-argon, Host: mgit@repsl-isle.us-east4-b.c.isleventures-vm.internal). Cross-referenced against live VM session logs (Cell eb815e95 on egg.isle.ventures), Founder_Institute_Sprint_repsl artifacts, and primary academic/legal sources. All phonetic ASR garbles, entity names, and temporal tense distinctions (past completed facts vs. present live states vs. future promises/proposals) have been verified and restored.] Opmatique solo contractor client intake & appointment-setting app * Richard Elieff (Opmatique) is developing a non-AI SaaS web application for solo contractors/handymen to replace fragmented messaging threads with structured client intake and appointment setting. * The current prototype handles structured client intake forms, price/spec adjustments, and calendar syncing without AI or on-the-job project management features (which remain potential future roadmap additions pending customer discovery interviews). * Keith Crandall offered to introduce tradespeople from his ongoing home renovation (managed by his wife) to help Richard Elieff conduct warm customer discovery interviews. * Kyle MacLean Smith suggested reaching solo contractors via warm introductions from recent homeowner clients (rather than cold outreach) and exploring modular API / MCP (Model Context Protocol) integrations for enterprise 'iPad contractor' workflows. seqSight genomic language model platform, patent applications, and pricing * Keith Crandall's company (seqSight, Inc.) is developing an AI-powered SaaS and API platform using compact, task-specific Genomic Language Models (GLMs) with Guided Tokenization to analyze DNA sequences for biotech companies and academic core labs. * The platform pairs seqLens (genomic language modeling) and statLens (hypothesis generation and statistical modeling across clinical/omics metadata) without requiring massive reference databases or bioinformatics pipelines. * Keith Crandall reported a pending non-provisional ('full') patent application on seqLens (converted in February from a prior provisional filing through George Washington University tech transfer) and a pending provisional patent application on statLens. * The current freemium/tiered pricing structure offers free sample uploads (up to 5 samples) alongside 3 SaaS pricing tiers and custom-negotiated enterprise API agreements, building on NSF-funded MIT open-source foundational publications. Enterprise data strategy & agentic AI orchestration consultancy * Amina Asim is building an enterprise data strategy consultancy and natural-language decision-support prototype to help data-rich organizations audit and orchestrate fragmented 1st-, 2nd-, and 3rd-party data before deploying AI agents. * Amina Asim is evaluating mid-market trade associations and 501(c)(3)/501(c)(6) non-profits in Washington, D.C., while navigating conflict-of-interest constraints with her current day job at Honda. * George McArdle and Kyle MacLean Smith recommended leveraging existing domain credibility for a beachhead market, offering 30-minute free consultations, and running reduced-price or charitable 501(c)(3) pilots. FI Working Group logistics (Cell eb815e95) and GovRes.ai federal job platform * The group confirmed Kyle MacLean Smith as Working Group President, opted to maintain a flexible entrepreneur-led discussion format, and scheduled recurring weekly meetings for Thursdays at 11:00 AM EST. * George McArdle (GovRes.ai) has built an LLM-powered web platform to help veterans and students tailor federal resumes for the 5-hour USAJOBS.gov application process. * George McArdle reported that institutional gatekeepers (veteran non-profits, university career centers, and D.C. government agencies) reflexively reject or ghost partnership and free-seminar offers due to vendor pitch fatigue. Product Architecture and Document Intake * George McArdle built the GovRes.ai platform around USAJOBS.gov's regimented layout to tailor applicant information to specific federal job requirements. * The system uses a general LLM to repackage user details for specific job listings. * The platform accepts information intake through PDF, JSON, and Word documents. * The application generates a clean, simple PDF for applicants to upload independently behind USAJOBS.gov login walls, since applicants must personally attest to resume accuracy under penalty of imprisonment (18 U.S.C. § 1001). User Acquisition and Outreach Strategies * Advisers warned against random LinkedIn ads because they produce poor click rates. * George established an affiliate program offering a 30% revenue share for signup links. * Amina suggested hosting free online webinars on Eventbrite or Luma to generate leads. * Kyle MacLean Smith noted potential non-profit reentry use cases teaching legal processes to incarcerated individuals (with LexClinic Director Anson Parker at the University of Virginia Library) who subsequently apply for paralegal jobs. Next steps * [The group] Prepare Upcoming Revenue Model Assignment: Prepare a 3-slide presentation explaining each startup's revenue model for the upcoming FI Revenue Sprint. * [Keith Crandall] Share Solo Contractor Contacts: Obtain list of solo tradespeople from his wife (project-managing their home renovation) to provide warm introductions for Richard Elieff's customer discovery interviews. * [George McArdle] Connect With Exited Genomics Founders: Introduce Keith Crandall to exited genomics founders in his angel network for workflow and commercialization guidance. * [Amina Asim] Introduce Neuro-Oncologist: Connect Keith Crandall with her brother-in-law (a neuro-oncologist and head of oncology at the University of Illinois) for potential cancer genomics research collaboration. * [Amina Asim] Contact American Chemical Society Lead: Reach out to her existing contact at the American Chemical Society (ACS) in Washington, D.C., to explore data strategy pilot opportunities. * [Keith Crandall] Introduce GWU Career Services: Look up and email George McArdle a warm introduction to the university-wide career advising center (and Milken Institute School of Public Health career services) at George Washington University. * [Kyle MacLean Smith] Consult Anson Parker on Government Document Standards: Ask LexClinic Director Anson Parker (UVA Library) about government job-application XML and business-process standards ahead of next week's call. * [Amina Asim] Share Federal Job-Seeker Facebook Group: Send George McArdle a link to the active women's federal job-seeker Facebook group. * [George McArdle] Evaluate Apollo.io for Granular Outreach: Investigate Apollo.io to identify and message individual federal job seekers directly on LinkedIn rather than running paid ads. * [Kyle MacLean Smith] Schedule Recurring Meeting & Coordinate Earlier via Agent: Set up the recurring Thursday 11:00 AM EST Legends Team calendar invite and circulate FI Sprint Working Group prompts via his Gemini 4 CLI agent in advance of each session. * [The group] Refine Problems: Eliminate 1 of the 3 core problems to focus on 2 for the upcoming sprint. * [The group] Share Assumptions: Share the key assumptions that the 3 core problems need for support. * [The group] Research Solutions: Identify and share known failed solutions in the pursuit areas. Want to see more? View the full notes Tip: You can always access your full notes from the left sidebar. You should review Gemini's notes to make sure they're accurate. Get tips and learn how Gemini takes notes Nov 5, 2026 Legends Meet #1 Invited kcrandall@seqsight.com george@govres.ai richard@opmatique.com david.huff82@gmail.com aminaasim@gmail.com Kyle Smith Attachments Legends Meet #1 Meeting records Transcript Recording Summary Founder Institute (FI Penn West) 'Legends Team' Working Group reviewed software products, customer discovery bottlenecks, and weekly scheduling consensus. [Agent Note — Forensic Audit Provenance (Cell 51a05033): This Google Doc was forensically audited and corrected in-place under the WeDo-JSON v1.3.0 protocol by REPSL Agent Cell 51a05033 (alias: call-transcript, Session UUID: 51a05033-ba49-412d-bfc8-a84d17867c13, Worker PID 12276 / Wrapper PID 12253 on pts/5, Window 5 inside GNU Screen 1413.repsl, Model: gemini-4-argon, Host: mgit@repsl-isle.us-east4-b.c.isleventures-vm.internal). Cross-referenced against live VM session logs (Cell eb815e95 on egg.isle.ventures), Founder_Institute_Sprint_repsl artifacts, and primary academic/legal sources. All phonetic ASR garbles, entity names, and temporal tense distinctions (past completed facts vs. present live states vs. future promises/proposals) have been verified and restored.] Web portal and sprints Participants addressed sprint deliverables and peer feedback requirements. Group agreed to review customer problems first. Product pitches and strategies Founders presented software tools for solo contractors and genomic analysis. Group discussed market positioning and outreach tactics. Meeting format and scheduling Participants agreed to maintain the current open format. Group established a recurring meeting schedule for Thursdays. Next steps * [The group] Prepare Upcoming Revenue Model Assignment: Prepare a 3-slide presentation explaining each startup's revenue model for the upcoming FI Revenue Sprint. * [Keith Crandall] Share Solo Contractor Contacts: Obtain list of solo tradespeople from his wife (project-managing their home renovation) to provide warm introductions for Richard Elieff's customer discovery interviews. * [George McArdle] Connect With Exited Genomics Founders: Introduce Keith Crandall to exited genomics founders in his angel network for workflow and commercialization guidance. * [Amina Asim] Introduce Neuro-Oncologist: Connect Keith Crandall with her brother-in-law (a neuro-oncologist and head of oncology at the University of Illinois) for potential cancer genomics research collaboration. * [Amina Asim] Contact American Chemical Society Lead: Reach out to her existing contact at the American Chemical Society (ACS) in Washington, D.C., to explore data strategy pilot opportunities. * [Keith Crandall] Introduce GWU Career Services: Look up and email George McArdle a warm introduction to the university-wide career advising center (and Milken Institute School of Public Health career services) at George Washington University. * [Kyle MacLean Smith] Consult Anson Parker on Government Document Standards: Ask LexClinic Director Anson Parker (UVA Library) about government job-application XML and business-process standards ahead of next week's call. * [Amina Asim] Share Federal Job-Seeker Facebook Group: Send George McArdle a link to the active women's federal job-seeker Facebook group. * [George McArdle] Evaluate Apollo.io for Granular Outreach: Investigate Apollo.io to identify and message individual federal job seekers directly on LinkedIn rather than running paid ads. * [Kyle MacLean Smith] Schedule Recurring Meeting & Coordinate Earlier via Agent: Set up the recurring Thursday 11:00 AM EST Legends Team calendar invite and circulate FI Sprint Working Group prompts via his Gemini 4 CLI agent in advance of each session. * [The group] Refine Problems: Eliminate 1 of the 3 core problems to focus on 2 for the upcoming sprint. * [The group] Share Assumptions: Share the key assumptions that the 3 core problems need for support. * [The group] Research Solutions: Identify and share known failed solutions in the pursuit areas. Details Web Portal and Peer Feedback Requirements: Keith Crandall noted confusion regarding how participants were supposed to submit peer feedback through the web portal for deliverables. Richard Elieff clarified that the group needed to complete last week's vision and mission sprint (set 12) regarding identifying core customer problems, alongside elements of the current customer development sprint (00:00:02) (00:53:32). The participants agreed to address last week's required question on sharing customer problems first before reviewing this week's sprint requirements (00:01:18). ________________ Richard Elieff - Target Audience and Solo Contractor Pain Points: Richard Elieff outlined the primary problem area for his Opmatique software product: verifying generic scheduling pain points for solo contractors, such as unlicensed handymen or independent licensed tradesmen operating as 'one-man bands' rather than as sophisticated multi-van businesses (00:02:11) (00:09:40). Richard Elieff explained that these solo contractors receive job requests and manage client communications through scattered messaging channels like Craigslist, resulting in fragmented details that are difficult to convert efficiently into confirmed appointments (00:03:11) (00:05:03). ________________ Richard Elieff - Product Scope and Appointment Setter Functionality: Richard Elieff described his Opmatique product as a SaaS web application designed to move clients out of messy messaging threads and into a structured client intake form (00:05:03). Richard Elieff emphasized that the tool operates strictly as an appointment setter without artificial intelligence, project management features, or labor and material cost tracking, though future roadmap additions might include estimation and invoicing support (00:05:52). Keith Crandall suggested conducting customer discovery interviews to verify if appointment creation aligns with actual contractor pain points (00:07:49). ________________ Kyle Smith - Back-Office Automation and Warm Introductions for Solo Contractors: Kyle MacLean Smith noted that while Richard Elieff is targeting solo contractors rather than enterprise "iPad contractors" (while suggesting modular API / Model Context Protocol integrations for future B2B expansion), cold-contacting independent tradespeople directly typically results in being ignored (00:10:19) (00:12:31). Kyle MacLean Smith recommended interviewing recent homeowner customers of solo contractors—starting with Keith Crandall and his wife—both to gather empirical workflow insights and to secure warm introductions to the contractors themselves (00:11:19). ________________ Keith Crandall - Genomic Analysis Platform and Specialized Language Models: Keith Crandall presented seqSight's SaaS and behind-the-firewall API platform, which uses AI-powered Genomic Language Models (seqLens and statLens) to help biotech companies (such as microbiome therapeutics firms tracking proprietary strains) and academic core labs detect novel pathogens and antimicrobial resistance (AMR) genes without requiring large reference databases or advanced bioinformatics pipelines (00:14:28) (00:31:17). Keith Crandall explained that unlike general LLMs like ChatGPT or Claude, or broad genomic foundation models like DNABERT and Evo 2, seqSight's compact, task-specific models preserve biological reading-frame information via Guided Tokenization (00:25:33). Keith Crandall mentioned completing a BARDA (Biomedical Advanced Research and Development Authority) accelerator (and partnering with Acorn Genetics and Oxford Nanopore for portable field deployment), having a pending non-provisional ('full') patent application on seqLens (converted in February from a prior provisional filing through George Washington University tech transfer) and a pending provisional patent application on statLens, and utilizing a 5-sample free trial plus 3 SaaS pricing tiers alongside custom-negotiated API contracts (00:17:48) (00:29:59). ________________ Kyle Smith and George McArdle - Market Positioning and Strategic Partnerships for Genomic Models: Kyle MacLean Smith advised Keith Crandall to highlight small-model precision—referencing a Frontier Syndicate Group presentation at Perkins Coie on agentic spikiness, the uncertainty principle, and July small-model research at Meta/NVIDIA—and to position seqLens as a specialized module within multimodal frontier stacks (00:28:49) (00:34:20). Keith Crandall noted that an NVIDIA genomics insider had invited him to speak right after the Evo 2 presenter at a UK National Health Service (NHS) conference in London. George McArdle and Amina Asim offered to connect Keith Crandall with exited genomics founders and Amina's brother-in-law, a neuro-oncologist and head of oncology at the University of Illinois (00:24:36). ________________ Amina Asim - Data Strategy Solutions and Customer Engagement Challenges: Amina Asim described her consultancy's focus on helping data-rich enterprises audit and orchestrate fragmented first-, second-, and third-party data assets before deploying AI agents (00:35:25). Amina Asim highlighted a core problem: prospective customers do not view data strategy as an issue, instead demanding immediate artificial intelligence agent deployment to cut costs, despite Amina Asim building a natural language query prototype for decision-makers (00:36:32) (00:38:46). Keith Crandall validated this fragmentation issue by sharing how George Washington University's grant administration data in myResearch remains split across 20 disconnected systems, unable to query grants where faculty serve as co-investigators rather than PIs (00:37:38). ________________ George McArdle and Kyle Smith - Establishing Credibility and Niche Markets for Data Strategy Consulting: George McArdle noted that Amina Asim faces a customer segmentation problem rather than a product problem, recommending that Amina Asim leverage past success as a beachhead or target specific niches like universities, trade associations near Washington, D.C., or automotive dealerships (00:39:56) (00:44:23). Amina Asim noted a conflict of interest with her current 5-year consulting role for Honda, preventing her from pursuing automotive manufacturers or dealerships until she transitions full-time to her business (00:45:16). Kyle MacLean Smith recommended Vitalik Buterin's writings on credibility/legitimacy ('The Most Scarce Resource Is Legitimacy') and suggested offering free 30-minute consultations and targeting 501(c)(3) charities (alongside 501(c)(6) trade associations) to build a track record without discounting commercial rates (00:47:12). ________________ Kyle Smith and Group - Meeting Format and Scheduling Consensus: Kyle MacLean Smith screen-shared his Gemini 4 CLI agent workspace (Cell eb815e95 on egg.isle.ventures, where he is building his Founder-as-a-VC / FaaV syndicate and shared knowledge repository) displaying the FI Working Group agenda and solicited feedback on whether to follow a rigid hotseat structure or maintain the current open-format discussion (00:50:18). The group agreed that the current blend of open entrepreneurial discussion and practical feedback was ideal. The participants reached a consensus to shift their recurring weekly meetings to Thursdays at 11:00 AM (00:52:28). Richard Elieff reviewed pending sprint deliverables, noting that teams must retroactively fulfill peer feedback requirements for both vision/mission and customer development sprints (00:53:32). ________________ George McArdle - Federal Job Application Tool and Institutional Resistance: George McArdle outlined his GovRes.ai software product designed to help veterans and students navigate the five-hour federal job application process on USAJOBS.gov. George McArdle reported a major obstacle: career counselors, veteran non-profits, and Washington, D.C. government agencies exhibit extreme resistance, refusing to pilot the tool, allow free 45-minute USAJOBS educational seminars, or respond to outreach despite public press releases lamenting the loss of federal employees (00:56:51). George McArdle attributed this barrier to institutional fatigue from frequent commercial pitches resulting in automatic rejections (00:57:57). ________________ Institutional and University Outreach Channels: George McArdle raised the problem that institutional gatekeepers block direct outreach to users and that student newspapers are no longer read. Keith Crandall noted that while students lack personal funds, university-wide career advising centers (such as the GWU Center for Career Services and the Milken Institute School of Public Health career office) have institutional budgets to pay for tools, after George McArdle noted that GWU's Trachtenberg School of Public Policy had previously rejected him three times. Keith Crandall and George McArdle agreed to exchange warm email introductions (00:59:18) (01:00:16). ________________ Online Community and Social Media Outreach Strategies: Amina Asim addressed the challenge of reaching job seekers by suggesting the use of alumni networks, university organizations, and Facebook groups where federal job seekers discuss employment hurdles. George McArdle noted difficulty in locating active Facebook groups for this purpose. In response, Amina Asim offered to send George McArdle a direct link to a specific all-women Facebook group focused on job hunting difficulties, while noting that admission restrictions might apply. ________________ Non-Profit Partnerships and Legal Education Use Cases: Kyle MacLean Smith introduced potential non-profit use cases, detailing his work with non-profits through the University of Virginia (UVA) to teach legal processes to incarcerated individuals so participants can apply for paralegal jobs upon reentry. Kyle MacLean Smith noted that this framework helps learners grasp the practical rationale behind their training while offering Socratic insights to improve the legal system, and highlighted his collaborator Anson Parker (Director of LexClinic at the UVA Library, misheard by ASR as 'Anton Clinic'), who drives standards in library sciences and legal process modeling (01:01:13) (01:02:40). ________________ Technical Architecture and Integration with USAJOBS: Kyle Smith asked about the degree of data science and technical innovation involved in the application. George McArdle explained that the system is built entirely around USAJOBS.gov, which features a highly regimented layout and requirements section, and that they utilize a general Large Language Model as an intermediary to repackage end-user data to directly address job requirements. Kyle Smith inquired about data formats such as XML and API calls, to which George McArdle responded that the intake system primarily processes PDF, JSON, and Word documents. ________________ Government Application Submission Constraints and PDF Generation: Kyle MacLean Smith questioned whether GovRes.ai handles direct government business processes or API submissions, offering to consult LexClinic Director Anson Parker (misheard by ASR as 'Anom') regarding government document standards before next week's call. George McArdle clarified that GovRes.ai generates a clean, simple PDF because the company is unaffiliated with the federal government and cannot submit applications directly due to USAJOBS.gov login walls and the statutory requirement (18 U.S.C. § 1001) that applicants personally attest under penalty of imprisonment that their resume information is true. Kyle MacLean Smith compared this model to arm's-length licensed intermediaries like patent agents, accountants, and attorneys (01:03:54) (01:05:01). ________________ Webinar and Social Media Advertising Strategies: Amina Asim proposed organizing free online webinars using Eventbrite or Luma to engage laid-off individuals and generate user leads, noting that George McArdle had not yet tried this approach. Amina Asim additionally recommended deploying social media advertising on platforms like Facebook to target specific demographic segments in the Washington, D.C., area before leaving the meeting. George McArdle acknowledged the recommendations for lead generation. ________________ Targeted Outreach via LinkedIn and Apollo.io: Richard Elieff asked George McArdle if they used LinkedIn to individually identify and contact unemployed people whose resumes display employment gaps in order to promote the application tool. George McArdle responded that while they operate an affiliate program offering a 30% revenue share for referral links, advisers explicitly warned them against running random LinkedIn ads due to exceptionally low click-through rates. Richard Elieff clarified they intended individual, person-level outreach rather than paid ads, prompting George McArdle to state that Apollo.io was recommended by others as a platform for this exact approach and that they intended to check it out. You should review Gemini's notes to make sure they're accurate. Get tips and learn how Gemini takes notes How is the quality of these specific notes? Take a short survey to let us know your feedback, including how helpful the notes were for your needs. Nov 5, 2026 Legends Meet #1 - Transcript 00:00:02 Richard Elieff: Keith, how you doing? Keith, can you hear me? Keith Crandall: I can. Yep. Richard Elieff: All right. Hi, Amina. Keith Crandall: Hey, Amina Asim: Hi, Keith Crandall: Amina. Amina Asim: morning. How's everyone? Richard Elieff: Good. Amina Asim: Well, you made it, Keith. I thought you were not going to make it. Keith Crandall: No, because we moved the time up to 10. I've got an alarm clock. So, I just started going in through the web portal for the deliverables, and there's a bunch of spots where we're supposed to get peer feedback, Richard Elieff: Correct. Amina Asim: Okay. Keith Crandall: but it's not obvious to me how we're supposed to get that because we Richard Elieff: Yeah. Oh, right. Keith Crandall: can't. Richard Elieff: Okay. So, the question comes down to Yeah. Do we get the feedback for just the Vision and Mission sprint now, which is a Set 12 question, or do we also delve into the current Customer Development sprint and that has a few areas where we need uh peer feedback. So maybe we do both today or just one today. 00:01:18 Richard Elieff: I don't know. Amina Asim: I think like the last one which was due uh on Tuesday or whatever it had like a question about peer feedback that you had to fill out to finalize Richard Elieff: Correct. Amina Asim: it. Richard Elieff: Right. Amina Asim: So assuming now we do next week's but we haven't done the work right. Keith Crandall: Yeah, that's well and it's not it's a lot of stuff too. Richard Elieff: Well, Keith Crandall: It's but it's it's not clear to me how to share that Richard Elieff: okay. Yeah, definitely. We should take care of last week's sprint, the one question about sharing the customer problem so we can check off the box and get credit for that sprint. Then I have started this week's sprint. So I could probably do ask you guys for some feedback on some of the stuff for this week's sprint so I can check off those boxes. Keith Crandall: Well, and that's this the the first question in this week's sprint is that last question from last week's anyway. It's the problems, right? Richard Elieff: Correct. 00:02:11 Keith Crandall: Identifying customer problems. Richard Elieff: Correct. Right. Keith Crandall: So, Richard Elieff: Share the core customer problems we're trying to verify. Keith Crandall: yeah. Richard Elieff: Right. Amina Asim: So we do last week's, then we stop. Keith Crandall: Well, it's the same. So, Richard, you want to go ahead and share your problems? Richard Elieff: Yeah. Okay. So, my three problems right now that the problem areas would be generic scheduling. I want to confirm that it's time consuming or confusing for them. Keith Crandall: is your target audience like general contractors. Just have the concept. Richard Elieff: No, that that's too high level up. Uh, so we're looking for solo contractors, one-man band, basically either a handyman that is unlicensed, does miscellaneous work, or a licensed tradesman like electrician, plumbing, flooring, roofer, whatever who's out on his own. So, Keith Crandall: All right. Richard Elieff: it's not a big business. It's just a guy in a truck and his dog, okay? Keith Crandall: Right. Richard Elieff: And he has to get his own business. He's—it's like hand-to-mouth. 00:03:11 Richard Elieff: You get a job this week, next week you do it. You got to go out and get another job this week. Next week you do it. Okay? So, it's not going to be a sophisticated business. Okay? So, and they're dealing through message channels. uh they advertise on maybe Craigslist or whatever inbound uh requests for hey I got a project are you interested what would it cost here you know what are the details that kind of stuff so they're they're building a business off a message platform so I want to verify that that's problematical then so the problem is that or I think the problem is that they'll have multiple requests coming in from different people maybe not all at the same time but different phases of trying to get these jobs then each job request has a message thread to it So his details all spread around. Then they got to convert all that into a confirmed project. They can, you know, get money from it. Keith Crandall: Right. Richard Elieff: So that's kind of the thing I'm looking at for my buyer persona. 00:04:06 Richard Elieff: Comments. Keith Crandall: I I have to say I'm deep into this world because I'm doing a whole Richard Elieff: Oh, how how so? Keith Crandall: home renovation. Richard Elieff: And you're hiring people to do it? Keith Crandall: or hire. Well, we have we have a general contractor, but my wife is deeply involved. She's the she's the painter because the painter wanted $50,000 just to do the painting of the house. So, my wife said, "Well, I can do the painting." But she's also been she's been basically the project manager. Um, so she's been finding the the plumber and the tile guy and the electrician and interacting with all these folks. Richard Elieff: Right. Okay. Keith Crandall: So we know this community quite well. Richard Elieff: Got it. I Okay. Hey, that's good. So, if you don't mind, uh maybe she can let me know who she's worked with and I can use them for my customer development interviewing. Keith Crandall: Yeah, we can try and get this place. Richard Elieff: Yeah. Oh, Keith Crandall: Y Yeah. 00:05:03 Richard Elieff: no. I got my own list, but it's all a cold call. So, if I can say that, Keith Crandall: Yeah. No, Richard Elieff: hey, Keith Crandall: we can get you a little more intro. Richard Elieff: I got referred to you by the Crandall family who you worked for. I want to talk to you about really, Keith Crandall: Yeah. Richard Elieff: you know, how you run your business. That'd be great. Okay. Okay. So, yeah, you got an idea of what I'm who I'm trying to pursue here. Keith Crandall: Yeah. Richard Elieff: Any feedback on these three different? Amina Asim: it like is it like an app or is it an app or what is Okay. Richard Elieff: It is an app. It's a SaaS software subscription app [Opmatique]. It's really a web application, and it works on a laptop or a phone, but they'll be using their phones basically. So, the idea is to get them out of the message channel, get them into a form on the app. They fill out the form, the client does basically comes into the uh handyman, they'll review it. 00:05:52 Richard Elieff: It may cause rejection. It may cause a confirmation, but that's how they kind of get their jobs. Amina Asim: So, are you building an are you building a marketplace? Richard Elieff: No, I'm I'm I'm not trying to match a client with a a tradesman. I'm just trying to convert messages that these tradesmen and handymen get into efficiently into appointments, jobs, basically. Keith Crandall: So does it automatically move information from a text stream into the Richard Elieff: No. Keith Crandall: app. Richard Elieff: No. Right. Right. There's no AI. So the point is to give the client a tool other than a message channel. You know, stop trying to create a message over, you know, five days of message back and forth and get all the information at once from the client in a form. The data comes to the handyman who will review the form and if there's some kind of fit, you can massage the data in the form and convert and confirm with the client, okay, I like this, but we have to change this and that. Here's a new price. 00:06:53 Richard Elieff: here's a different spec, here's a new time and date. So, it eliminates the appointment building process in a message form through a thread and it puts into input form, client intake form, Keith Crandall: And is it just for getting the job or does it continue through the job? Like, you know, with our tile guy, we're we're consistently changing our minds about tile type and grout Richard Elieff: right, Keith Crandall: colors and, Richard Elieff: right, right, Keith Crandall: you know, Richard Elieff: right. Okay. Keith Crandall: trying to figure out when he's going to show up and all those type of things. Richard Elieff: Yep. Yeah. Yeah. Okay. Keith Crandall: So, Richard Elieff: Good. Good point. So, yeah. Once the um handyman confirms the appointment and the details, it can go and sync to his calendar and then you can monitor the status of the appointment, whether it's uh rescheduled, whether it's you the confirmation is pending, whether it's completed. But you really can't track the job details yet. You'd have to probably set up some kind of separate system. 00:07:49 Richard Elieff: So it is focused on appointment creation but it's not focused on project management. Now I envision that down the road if this gets traction and people like it and I got to grow the functionality. I think what they may want is more of um support on the front end which is hey help me quickly and efficiently estimate the price of the job because that takes time and then help me you know do billing and invoicing because they got to get their money when they're done. Keith Crandall: Yeah. And that that comes down to actually doing those customer interviews, customer discovery interviews, Richard Elieff: Correct. Keith Crandall: and talking to people and figure out whether the appointment concept, if that's the pain point versus the tracking of the details. once you're on the job versus keeping track of your 20,000 text messages Richard Elieff: Correct. Right. Keith Crandall: on five different jobs, Richard Elieff: Right. So, right. It doesn't track man-hours spent. It doesn't track, you know, cost of goods that you've got to buy—all these paint buckets and all this square footage of tile and all these roof shingles. 00:08:50 Richard Elieff: It doesn't do that. So, it's just strictly right now an appointment setter. Keith Crandall: right? I'd say do the, you know, do those customer discovery interviews and figure out whether their pain points align with this. And if not, maybe you pivot a little bit and shift to wherever their pain points are, Richard Elieff: That's correct. Keith Crandall: right? Richard Elieff: Right. So, these were the pain points I just mentioned, the three that I'm going to pursue. Any issues you guys have? Kyle Smith: Well, I my thought with Keith, I think your insights were really great because uh Richard, you're trying to go directly to the contractors and the Richard Elieff: Yeah. Yeah. Yeah. They're not contractors again. Well, I'm going for the Yeah, Kyle Smith: Okay. Richard Elieff: go ahead. Go ahead. Keith Crandall: Sounds Kyle Smith: Please let me know the exact end customer. That's important. That's not an aside. Richard Elieff: Yeah. Oh, right. Okay. Kyle Smith: That's that's what we're doing. Richard Elieff: So, yeah, it's not a sophisticated tool. 00:09:40 Richard Elieff: Again, there's no project management involved. It's not for teams. It's not for a company that has a bunch of technicians that dispatch, each have their own van, that kind of stuff. It's for a solo dude. It's pickup truck and a dog. They go just, you know, they get their jobs week by week, day by day. They run out and do a few hundred bucks of work or a few thousand bucks work and they're trying to drum up more work week by week. So, it's very informal. They probably take payments by Venmo or cash or whatever like that, digital cash. So, it's um these guys might be operating on paper and pencil for all I know right now. So, Kyle Smith: Okay. So, for me that's a that's a solo. Richard Elieff: I don't want to bombard them with all this crazy. Kyle Smith: Yeah, for me that's a solo contractor. So, that's what I mean by contractor. Richard Elieff: Okay. Kyle Smith: A solo contractor. Richard Elieff: Yeah. Okay. 00:10:19 Kyle Smith: I do know that there's a meme that is iPad contractor that is for uh enterprise that have many people who work for them. So, if you look up 'iPad contractor' on YouTube, that might be helpful, because you're describing a solo contractor, someone that might have a pass-through LLC. [Agent Note — 'iPad Contractor' vs. Solo Tradesman: Refers to the industry trope of private-equity-backed residential service roll-ups whose commissioned sales technicians arrive with iPads running enterprise field-service and tiered-financing software (such as ServiceTitan), in contrast to Richard Elieff's target beachhead of independent 'man-in-a-truck' solo tradespeople.] They might just be doing sole proprietorship. Uh and and you're just saying, "Hey, every month, if you've got a credit card, pay me for for this." And this uh makes your conversion uh less expensive for you, right? For for the customer. That's your basic uh premise is that it's going to be less expensive for them. It's going to be less of a headache for them to do their work for their clients, the back office. You're you're automating their back office, and that's why they're going to give you however x a month, right? Richard Elieff: right? Exactly. Yeah. Well, Kyle Smith: Okay. Richard Elieff: yes. Kyle Smith: Yeah. Richard Elieff: uh automate part of the back office. Kyle Smith: I know not all of it. 00:11:19 Kyle Smith: I know not all of it. Yeah. Yeah. None of us can do all of it. Whoever is saying panacea are lying. But uh um Okay. So that's great. I think the thing about the iPad contractors, it'd be interesting for you to consider what you're doing is a module that connects with their operations. And so that's APIs. I'm a big fan of monetizing uh B2B with with APIs with the um business processing you're doing because you know it's not part of the complete solution and you know in the way it's not so you already understand it as a module. So if you connect an API—MCP [Model Context Protocol], for instance, a very modern API—then that probably could open up a customer segment but but I importantly for what you're doing now the solo person for this week because because you're already focused on that persona and you've done a lot of research on that I believe Keith has opened up a really important avenue for you to be able to actually access them and uh it's also for you to validate that you're actually the value-add you're hoping to be; he's also surfacing where you need to do that kind of empirical process. 00:12:31 Kyle Smith: So, him and other people such as him, people who've had a a job recently done by a solo contractor, Richard Elieff: Right. Kyle Smith: not an iPad contractor, reach out to them and then they can get you to in contact with their contractors, but as you pass through, you also can pick up important knowledge from them. So to me, that's the way you access the persona that you do need to interview, which is the solo contractor, the person who's going to buy your SaaS. But yeah, you you actually go through Keith and other people to get to them because you're saying you go directly and it just blocks, right? They just get blocked or you just get blocked, right? Richard Elieff: between an iPad contractor and a solo contractor. Kyle Smith: Um yeah, if you contact a solo contractor cold turkey, they ignore you. Richard Elieff: Right. Kyle Smith: But if you the the the premise here is if you go through recent customers of them such as Keith, Richard Elieff: Right. Kyle Smith: then they're going to give you the time of day. Richard Elieff: Yeah. 00:13:25 Richard Elieff: Sure. Right. Kyle Smith: So you contact a Keith type persona, Richard Elieff: Right. Kyle Smith: Keith specifically, as you just did just now. So you repeat that relationship with someone else such as Keith who just had a a job done by independent contractor. And then you learn what Keith just said. He said some really important information to you in that interview. And then that gets you the warm intro to the contractor. So you should know have that knowledge base anyway for you to be the value-add you're saying you're going to be, to have traction and grow legitimately as a product—people know it's actually solving the pain point—and then it'll also give you a warm intro you need to interview to see that there's somebody to pay for you solving that pain point. So, it's a little extra step, but since you're not able to directly contact the contractor right now, uh I believe it's a um a smart um two-step process to to get you where you need to go. Richard Elieff: Okay, thanks. 00:14:28 Kyle Smith: Uh, George, you're on mute. Amina Asim: Yeah. George McArdle: Uh, hey, sorry I'm late. Amina Asim: Hi. George McArdle: Um, Kyle, if you want to, you can boot my notetaker. It automatically joins and joins anything I I sign up to. Kyle Smith: Okay. Uh, I'm taking notes for all of us and I'll share them. So, if if you don't want to duplicate, that's totally fine. George McArdle: Oh, Kyle Smith: Oh. Keith Crandall: Oh, George McArdle: there we go. Keith Crandall: I can go next because I developed a few problems. So again, my my company [seqSight, Inc.] is developing AI-powered genomic analysis platforms to help bioinformatics teams at diagnostic and biotech companies detect novel pathogens and antimicrobial resistance genes that database tools miss. [Agent Note — seqSight, Inc. (seqsight.com): Co-founded by Dr. Keith A. Crandall (Director of the GWU Computational Biology Institute) and Dr. Ali Rahnavard ('DNALingua DBA seqSight'), commercializing task-specific Genomic Language Models (GLMs) with Guided Tokenization ('seqLens', 'resLens', 'microCafe') and closed-loop statistical modeling ('statLens').] So we're doing a software as a service and and AI approach to analyzing DNA sequence data. DNA is just a different kind of language. Um, and we've developed a different kind of large language model to to uh gain inference from DNA sequence data. So, so our our core uh clients are biotech companies, not big pharma really at the moment, but more biotech companies who do a lot of DNA sequence analysis and core sequencing labs that do a lot of DNA sequence generation for clients who don't really have the chops to analyze the data because it takes a lot of bioinformatics skill and expertise to do the data analysis. 00:16:04 Keith Crandall: is so we've developed a platform that allows you to just upload the data and march through the analysis. So our problem one is for microbiome therapeutic companies—they can't reliably track their own strains, right? So a lot of these biotech companies have their own bacterial strains for proprietary reasons like you know Botox and things like that. That's a you know that's a their own bacterial strain. uh beer makers have their own proprietary yeast strains and things like that and they want to do comparisons with other things out there but they don't necessarily want to um provide their data to outside sources. So we have an API approach so they can pull our tool into their environment behind their firewall and do analyses um and get insights without having the nice thing about the the genomic language models you don't have to set up huge databases reference databases which is what you normally would need to do. So our second problem is um folks who are looking for drug resistance variants right there are a lot of antimicrobial uh drug resistance is a huge thing in human health and identifying those variants is very difficult uh and time consuming from DNA sequence data takes a lot of informatics that a lot of um clinics and and uh biotech companies don't have Um it also allows um 00:17:48 Keith Crandall: well we our approach again is takes no Richard Elieff: Thank you. Keith Crandall: bioinformatics skills, no large databases, and it can identify new drug-resistant variants, right? The beauty of the genomic language models is it can identify stuff that people haven't seen before. So this really opens the door. We just did a BARDA [Biomedical Advanced Research and Development Authority] accelerator, and it opens the door to government contracting through things like BARDA and FDA for identifying novel pathogens and things like that. And then the third problem are academic core labs um that that produce a lot of data for their clients who are typically biotech companies or academics at medical school, school of public health. Uh and those people that are seeking their services typically don't have um strong capacity in analyzing data. So uh this provides a a unified platform a uniform platform to do all this stuff uh to integrate their clinical data and the genomics data to do analyses. Um, and we're thinking about um well, I'm I'm thinking about prioritizing the the antimicrobial infection in the academic core labs just because I know the academic core lab environment and uh and I I know how to reach out to those people and I know their pain points pretty well. 00:19:23 Keith Crandall: in the antimicrobial infection space, because we've done some work with the FDA and we've been through this BARDA accelerator, you know, we have more traction there than with the biotech industry. So, that's kind of what we're thinking at the moment. I know this stuff is far afield from what you all are doing. Amina Asim: How does genomics uh or the data that you're providing fit in with like all the other types of research that's going on like cancer cells and everything? Does it is it aligned with that or is it separate? Keith Crandall: Yeah. Yeah. So, so one of the applications of our genomic language models would be to identify cancer variants from DNA sequence from a sample from a patient, right? We don't want to get down that clinical road yet because that's a higher hill to climb, right? You need a lot more validation. You need FDA approval. We've got a collaborator—he's on our Science Advisory Board—who's a pathologist at the GW Hospital in DC where we can we're working with him to get data from samples where he's doing the normal diagnostics whatever he does right and then getting permission to do sequencing on those same samples and then and then produce our results whether he's looking for infectious disease or cancer variants or other disease variants that anything that smacks 00:20:55 Keith Crandall: of of um where where DNA is is going to be useful in insights. Then we'll get duplicate samples, sequence, Amina Asim: Oops. Keith Crandall: compare results, and then track the time it takes us to do our end of things versus the time that it takes him to do his end of things. the the the quality of the diagnostics and the the uh insights of the information that's provided back and how that how that might change uh a patient's experience hopefully for the better and more accelerated and reduce uh the hospital cost because in the end you need to do all of those to get some new technology into a into a a hospital setting. So, so that's that's something that we've got, you know, working on on the side, but that's kind of we need to get revenue in before that happens. Amina Asim: What's the difference between like making your data available which I understand like it's like really from what you described yesterday. When did we make it? It's been it's been a long week. uh like getting your data accessible versus just like the analytical layer that you're talking about which is the meat of it. 00:22:15 Keith Crandall: So I think it's for a lot of people it's it's the analytical layer that's difficult uh and and bioinformaticians are expensive and they don't necessarily know the breadth of the analytics and statistics that we combine into one platform. Right? Some bioinformaticians can do this part, some can do this part, but it's very hard to put it all together and they tend to put it all together every time they do an analysis as opposed to having it all together in one platform where you just upload your data. In fact, when we were doing customer discovery through the I-Corps [NSF Innovation Corps], we talked to a lot of bioinformaticians at biotech companies, and they love the concept because it saves them time, right? Because they don't have to build the pipelines; they can do the analyses and get results back faster to other people in the company. Amina Asim: Oh no. Keith Crandall: So, they're excited about that concept as well. George McArdle: Keith, could I could I throw in a question? Have Have you talked to other startups um that aren't doing exactly what you're doing, but are doing kind of the piece on either side? 00:23:20 Keith Crandall: Yeah. In fact, at this BARDA—the BARDA accelerator had a sort of graduation pitch event in Chicago, and I met another company called—I got the guy's card—Acorn Genetics, that does the front-end sample prep before sequencing on a particular sequencer that's portable, and his front-end prep is portable. And BARDA is all about deployment, right? We need to be able to deploy technologies in the field wherever we go. If there's an Ebola outbreak in the Congo, we don't want to be collecting a bunch of Ebola samples and then bringing them back to Frederick, Maryland [Fort Detrick / USAMRIID], right? And doing stuff. We want to be able to do the analysis right there and right then, because everything else takes weeks or months instead of a few hours. So we're actually working with Acorn [Acorn Genetics] to put the front-end sample processing, Oxford Nanopore [Oxford Nanopore Technologies] as an established portable sequencing technology, and then our analytical capacity on the back end. [Agent Note — Portable Field Genomics Stack: Acorn Genetics (a BARDA Accelerator Network I-CREATE MedTech showcase company) develops portable molecular sample-prep and diagnostic hardware ('SENTRI' / 'AcornLab'), combined here with Oxford Nanopore portable sequencing and seqSight's lightweight 'seqLens' backend for in-situ pathogen and AMR detection without cloud or large reference-database dependencies.] 00:24:36 Keith Crandall: So, you package these three things and you can deploy anywhere in the world and do it in situ, right? George McArdle: that that uh that that seems like a my angel group has invested in a few genomics things just because uh one of the partners is really invested in that space and and one of the things we had heard back was that you know people say oh this is great but what I really need is this plus this plus Keith Crandall: right? George McArdle: this so that an entire workflow is just off my plate. Um if you want to hit me up afterward, I'm not uh okay. Keith Crandall: Yes, I do. George McArdle: Yeah, I I'm not quite competent enough to advise on this, but um I know a couple of guys who I I think would like a couple exited founders who I I think uh were in the genomics space might might be really worth talking to. Keith Crandall: Yeah, sounds great. George McArdle: Great. Amina Asim: Uh, I think my brother-in-law is a neuro-oncologist at the University of Illinois. So, he does a lot of cancer research. 00:25:33 Amina Asim: So, I can put you in touch with him if you like. He's the head of the oncology. Keith Crandall: Yeah, that'd be great. Amina Asim: Uh, Keith Crandall: I'm one of my sons is actually a PhD student at University of Illinois although in Amina Asim: oh yeah. Keith Crandall: philosophy Amina Asim: Yeah. He's a neurooncologist, but he he's doing like cancer research. And you know I think the the question that I get to is like you know why can't you just plug stuff into Claude? That is, I think, the biggest Keith Crandall: and of course the reason is because Claude and ChatGPT are general English large language models. Amina Asim: enterprises. Keith Crandall: One, some people don't want to feed them their data. Two, Amina Asim: Yeah. Keith Crandall: they they treat DNA sequence data like like English language text and it's not. Amina Asim: Mhm. Keith Crandall: It's a completely different language with completely different meanings. So our genomic language models understand that and and we have developed our own models that take into account the biological realities including unique model selection and unique uh what we call guided tokenization. 00:26:36 Keith Crandall: So when we chop the words up in the genomic language, we do it in a way that preserves the biological information, whereas the tokenization in Claude and ChatGPT does not, right? So we get much better insights. Amina Asim: I think that's super cool and interesting that it would be its own type of LLM. Um, Keith Crandall: Yep. Amina Asim: yeah, I Keith Crandall: And there are other large—there's DNABERT and Evo 2 now—that are broad language models for DNA trained on all DNA sequence data. [Agent Note — DNABERT & Evo 2 vs. seqLens: DNABERT (Ji et al., Bioinformatics 2021) and Evo 2 (Arc Institute & NVIDIA, Nature March 2026; a 40B-parameter StripedHyena 2 foundation model trained on 9.3 trillion nucleotides across OpenGenome2) are broad foundation models trained across all genomic corpora, whereas seqSight's 'seqLens' (Baghbanzadeh, Mann, Crandall & Rahnavard, Molecular Biology and Evolution, July 2026) trains compact, task-specific models using biologically informed Guided Tokenization (GT).] Amina Asim: Okay. Keith Crandall: But the problem with those is they're huge language models and and they're trained on everything. And when we compare our results to those models, we do much better because we we train specific models for specific tasks. So if we want cancer variants, we train on cancer variants. And if we want microbiome, we train on micro microbial genomics, right? If we want antibiotic resistance, we train on antibiotic resistance. So our models are much smaller. They're much more compact. 00:27:47 Keith Crandall: They're much more efficient. Consequently, they're much more sustainable, but also uh they're just better, right? You get more accurate results. Amina Asim: And then they're deployed in that particular environment, which is, I think, a big selling point. Keith Crandall: Yep. Richard Elieff: Keith, just out of curiosity, what is your foundation? These small LLMs you're using as a foundation that you're then customizing and training—can you disclose that? Keith Crandall: Well, it's our software called seqLens that we have a publication on. It's an open-source publication. It's our targeted genomic language model. So, it's all described in the paper, yeah. Richard Elieff: So, okay. So, you built your own specialized LLM that understands genomic language rather than English Keith Crandall: Yeah. Yep. Yep. Richard Elieff: text. Is that right? Keith Crandall: Right. Exactly. Richard Elieff: Ah, wow. Okay. Impressive. Okay. Okay. Kyle Smith: Actually, Amina Asim: I can I go now? 00:28:49 Amina Asim: I I could use some help. So, I think one of Oh, Kyle Smith: can can I sorry, Amina Asim: yeah. Kyle Smith: I just like to give Keith Keith some comments. Amina Asim: Sorry. Go on. Kyle Smith: Uh that that's really great, Keith. Um there's uh I there are groups I know definitely in Silicon Valley that are multimodal harness people. Some of them do the self uh improving harnesses and they really understand patents. So, I'm a little bit chagrined on—or not chagrined, a little bit nervous about—this suggestion here, because I don't know if they're being the best stewards on patent management, but I do think that market in in general the multimodal really sophisticated frontier teams uh would understand and be much better able to leverage your model within the their stack. So, uh, for instance, finding patentable subject matter and then publishing that. They they also work with patent agents, so they've got that whole vertical figured out. And I mean, I chat with the Wolfram Research folks as well, and those are, I guess you'd call it, hard 00:29:59 Keith Crandall: Yeah. Kyle Smith: AI—but yeah, the variety of models and multimodals is definitely important as we go towards AGI. So have you thought of that as you being a module? And I think for me what's important to understand is how your API works, because I understood it's an API, but then it's SaaS, but also people are downloading open source—or maybe I'm wrong on that last part. Keith Crandall: Well, we we always the way we do things is we always publish the because it's the the foundational models are kind of funded by by federal grant dollars, NSF. Um so we publish open source and publish uh we have students working on stuff and they publish their open source but then but then we take that um as sort of foundational work and then re we re well we redo it in a more sophisticated way um for the company. And we have a pending patent application on our seqLens concept—our genomic language model that includes the tokenization and model selection and all these things. 00:31:17 Keith Crandall: So we did a provisional patent application a year ago in February, and then this past February it moved to a full [non-provisional] patent application. So that's as part of a tech transfer arrangement with GW [George Washington University], right? So that's already in the works. We have another patent pending—a provisional patent application pending—on the statistical end of things. So we have seqLens, that's the genomic language model to do all the genomic insight, and then we have statLens, that's specifically trained for doing statistics and hypothesis generation and testing, so that when you get—you know, it's one thing to analyze the genomic data and get the insights, but then you always want to see how those relate to all the clinical features and demographic features that you have. So we can bring in a bunch of other metadata, even other omics kinds of data—metabolomics data and lipidomics data and all that—and do sophisticated statistical analyses without being a statistician, right? [Agent Note — Temporal Tense Audit on Patent Status: Both the non-provisional 'seqLens' filing (converted in February from a prior provisional application through GWU tech transfer) and the provisional 'statLens' filing are pending patent applications ('in the works'), not yet issued patents.] 00:32:27 Keith Crandall: So, we have that incorporated into the platform as well and found both of those Kyle Smith: Okay. Keith Crandall: going Kyle Smith: So I would then guess it's a freemium model—so quite a bit of it is free, but then people do pay for the API for the Keith Crandall: Yeah. So, Kyle Smith: professional Keith Crandall: so the current pricing structure and that's I'm hoping that when we get around to talking about things like that uh to get more feedback on this because I really don't have any clue on this but uh our current pricing structure is you know you can go to our web platform form and and upload, you know, five samples and do some analyses and get a feel for the platform and see if it looks useful to you. And then and then there are like three pricing tiers. Uh and then API is kind of um give us a call and we'll negotiate and see what you're up to. Kyle Smith: Okay. And you do have MIT [open-source licensing] for at least a good chunk of it as well. 00:33:20 Keith Crandall: Yeah. Kyle Smith: Uh and then the other one I wanted to say Oh, uh it's it's escaping me right now. Uh Richard, I saw that you had something you wanted to say to Keith. I think you you looked excited. Okay. Well, uh, Richard Elieff: Yeah. Kyle Smith: for So, what I wanted to Yeah. Richard Elieff: Hey. Kyle Smith: Okay. Good. Good. Richard Elieff: Okay. Kyle Smith: Please. Richard Elieff: Yeah. Real quickly, Kyle Smith: Yeah. Richard Elieff: we will talk about pricing. In fact, it's this week's assignment. There's a set of issues about a revenue model. Keith Crandall: Oh, Richard Elieff: Got to develop a three slide, Keith Crandall: we got that. Richard Elieff: you know, whatever to explain next week. So, yeah, it's coming up. Kyle Smith: Okay. So, final thing I just wanted to say about your customer discovery is specifically Meta—there's a Frontier Syndicate Group that I'm part of, and they gave a presentation at Perkins Coie about recognizing small models are more precise in many applications, the spikiness of agents, and so what we think is really underlying it is the uncertainty principle. [Agent Note — Small Language Models, Perkins Coie Frontier Syndicate Presentation & Uncertainty Principle: References the Frontier Syndicate Group presentation at Perkins Coie LLP alongside July 2025/2026 research on small vs. frontier agentic models (cf. Belcak et al., NVIDIA Research, 'Small Language Models are the Future of Agentic AI', arXiv:2506.02153, and Meta's Llama / Muse Glimmer small-scale agentic distillation), connected in Isle Ventures' thesis to conjugate precision/context trade-offs analogous to Heisenberg's uncertainty principle (cf. Prof. Philip Moriarty, 'When the Uncertainty Principle Goes to 11').] 00:34:20 Kyle Smith: So think about Meta and that group, because they're very primed for it—they did a paper in July. At the very least, you can cite that paper to explain how multimodal is the future and you've got this model Keith Crandall: Well, Kyle Smith: that's amazing. Keith Crandall: we also have an insider at NVIDIA who works in genomics, and Kyle Smith: Great. Keith Crandall: and that's his thesis—he invited me to go to this National Health Service [NHS] conference in London to give a talk, and I was right after—strategically right after—the Evo 2 guy, that's the big one. And in fact, I have this slide that's this big whale and then a bunch of little fish, right? And Evo 2 is the big whale trying to do everything, and we're the bunch of little Kyle Smith: Mhm. Keith Crandall: fish that do things much more efficiently and much more accurately than the big whale. So yep, and that's ultimately what we hope, right, is that we just get bought out by somebody like NVIDIA and be done with Kyle Smith: Thanks. Keith Crandall: it. 00:35:25 George McArdle: Uh, cool. Do we want to jump along? Keith Crandall: Yeah. Have you have you George worked or Kyle or because Amina you haven't had a chance to get into this yet, right? Amina Asim: Yeah, I I would love some like brainstorming and whatever help because like I feel like the the my interest is in the data strategy part which is that you've got like most of the companies that we work with would be like data rich they have a lot of first, second, third party data that they're not utilizing. like efficiently yet. My problem is that the customer doesn't see this as a problem and I've only encountered it because I've been in the data environments and then I'm like, well, no one's using all of this. No one's planning this out. No one's, you know, talking to each other and then having played that role seeing like people save or make millions from like a good data strategy. But I think it's not like approached that way by most companies. So my my problem is that the customer doesn't really see it as a problem. 00:36:32 Amina Asim: The the problems that the customer is seeing is like we want to build AI agents or we want to invest in AI and we want to like you know fire a few people so that we can decrease our costs etc. And so like when we talk when I talk to people they're like oh can you just build us AI agents and what I'm trying to sell is that no before you before you build AI agents you got to like figure out all of the data that you have you got to audit all of it all the different sources align it to your business objectives then build the agent because like otherwise you're not going to make the best of it. So my biggest problem is that the customer I think doesn't really see it as a problem because they've got their analysts and they've got their business people and they don't talk to each other but they've hired people to do all of this for them and they get like their powerpoints which help them make decisions. Keith Crandall: do you have do you have users who want this integrated data because I'm I mean I'm at a university right that has a lot of data and and they've implemented systems 00:37:38 Amina Asim: Yeah. Keith Crandall: to manage the data to make it easier to extract data for what they want, and yet—I mean, one basic thing is, this is classic, right? We have this system called myResearch [GWU's grant administration portal] that we submit grant applications with. Amina Asim: Mhm. Keith Crandall: The the goal of this was that the university admin could extract all the information they wanted on grant submissions. And yet I just got an email two weeks ago asking me for all the grants I submitted where I wasn't PI where I was a co- co-investigator because somehow the system doesn't track that information. Amina Asim: Uh-huh. Keith Crandall: Right? So, and yet the research office knows all its information because they track every grant, right? They process every grant. So, they know all the key personnel on that grant. They know all the positions, all they know all the budgets. And yet, nobody seems to be able to find this information. And clearly, somebody wants it. and and and then they come back to individuals and say, "Can you you know?" So, 00:38:46 Keith Crandall: in this case, it's my finance manager who's the one trying to find all this information and can't access it because it's all, like you said, in 20 different places across the system [ASR: 'across Steam']. Amina Asim: It's completely fragmented. There's a lot of systems in place. There's a lot of data that no one's looking at or using and they don't even have to. But for like decision makers, something like this comes up exactly like this and then they're like find trying to find out who's going to give them that information. So our pro we built a prototype too which basically gives the decision maker like natural language. they can just type in their question and all once you've done all the data orchestration and agentic kind of talking in the background then they can just get an answer and the source of the answer. So that would be like the ultimate kind of subscription part of this. But George McArdle: Um, Amina, can I um, Amina Asim: yeah. George McArdle: it sounds from the outside that you don't really have a product problem—like, you kind of know that there are people out there, and disparate data is, you know, something everyone kind of knows AI can help with. 00:39:56 Amina Asim: Yeah. George McArdle: Um, it sounds like you kind of have a customer problem at the moment in that you're you're you're seeing your your your tool can be applied to any kind of customer, Amina Asim: Yeah. George McArdle: be applied to a university, it could be applied to a mid-sized corporation, a huge-sized corporation. That kind of generic approach works for like Google; you are essentially a consultancy. Amina Asim: Mhm. George McArdle: Um and the usual way that a consultancy gets customers is through referrals and uh so you know you've said you've had a Amina Asim: Mhm. George McArdle: lot of success in the past through you know doing this through a larger organization for other organizations. Um like the the two suggestions I I would go through is one reach back into that network and then kind of say hey we're doing the same thing you saw what I did before. this is going to be 30% cheaper. Um, I would love to work with you on this. If not, if you know anyone, you know, I I can help them is kind of you need the warm intro at the moment. 00:41:08 Amina Asim: Mhm. George McArdle: Um, the second one is the other approach there is you can try to find a niche Amina Asim: Yeah. George McArdle: like Keith just described like a very real pain point for universities. Amina Asim: Mhm. George McArdle: So you could become the university person who like swoops in, has a playbook, gets in, gets out after 3 months, Amina Asim: Mhm. George McArdle: and they immediately see a 10%, you know, increase, and then you say, "Okay, cool. We have a ongoing subscription that will kind of keep these like things going for you." But but it it sounds like you kind of need a you need a beach head. Um, right now you're sort of trying to talk to anyone who will Amina Asim: Well, cuz I don't have any customers. I was going to focus on like media companies where then the only people who showed interest were like automotive and agencies because I have background working with automotive data and agency data. So then I'm like okay well who you know who am I to say no you George McArdle: No, Amina Asim: know let me follow this lead. 00:42:09 George McArdle: that that's like a Yeah. Amina Asim: Yeah. George McArdle: I is you using using the thing where you have a track record of success as your beach head Amina Asim: Yeah. George McArdle: market unless you've looked at the market and been like no there's nothing there. Um like I if you can if you can start there get revenue get success Amina Asim: All right. George McArdle: stories and then present as hey I have landed this niche by the way we've just opened up to where we're willing to work with Ford you Amina Asim: Mhm. Mhm. Yeah. George McArdle: know um and and um I I I got to agree with the guys who um uh were giving you uh uh some advice the other day is you might have to start off with a free pilot. Uh I I I it just um unless unless again it's somebody who you've worked with before who you've shown success to. Amina Asim: Yeah. George McArdle: Um is you know and it can be a very very it can be a paid pilot and it can be very very cheap. 00:43:14 George McArdle: Um but you know for for anyone to give you any money they probably need to know that you're going to deliver them success. Amina Asim: Yeah, I think the advice I've gotten is to start giving free services like and con consultations. Um, and then that might lead to something. George McArdle: Consultations are a given. Like you should definitely be giving out free consultations, Amina Asim: Yeah. George McArdle: but um you know pilots people disagree on whether or not they should be free or not. Amina Asim: Yeah, I I think I would just do a reduced price pilot, but not like, you know, the full price. But thank you. Yeah, I think that's what I'm struggling with is like trying to find a niche that I can focus on. I was thinking about associations. Like there's like these massive associations all over DC and I bet they have like s*** tons of data but I don't know if they have money. George McArdle: That would be I mean that that would be really good recon. Amina Asim: Yeah. And it would be interesting too cuz they have like research market. 00:44:23 Amina Asim: They are old school. They probably need help. George McArdle: Um, do you have any connections to any kind of association of auto type people? Amina Asim: No, but I can, you know, start like trying to find them. Like if I can focus, then I can find the the customer and find where they're at and stalk them basically, you know. But like I think because I'm not focusing, I go to all these events and I meet random people from like random industries and then someone gets interested and then I get distracted in that direction basically chasing the lead. Keith Crandall: Hey, George McArdle: Yeah. Keith Crandall: American Education Association [NEA / AERA], American Chemical Society [ACS] are right near where we meet [in downtown Washington, D.C. near GWU]. Amina Asim: Yeah, I was thinking about the chemical association for some reason. I met someone from there and they got interested as well. So I could just reach out to her. Richard Elieff: Amina, quick question: when you say automotive background and data or whatever, what kind of automotive business is that? Is that like a manufacturer? 00:45:16 Richard Elieff: Is it a dealership? Amina Asim: It's Honda. Honda. I I've been working for Honda for 5 years and I the Richard Elieff: Honda. Honda the corporation, right? Amina Asim: the company. Yeah. Richard Elieff: Well, how about maybe Honda dealerships? Amina Asim: Well, Richard Elieff: There's many more of them. Amina Asim: it's a conflict of interest right now. So, I can't go after them. that I like that's why I know there's value in this cuz I built all of Richard Elieff: Okay. Amina Asim: this Richard Elieff: some some other brand dealership, you know, Toyota or, you know, you'd be dealing with the local entrepreneur, not the corporate bureaucracy, and they could make decisions locally. You could probably get your pilots running quicker at a smaller scale. And they have a bunch of data they're trying to manage to run an automotive dealership. Amina Asim: honestly like I know the dealership data it's uh it's just not something I can pursue right now. I think once I have some income coming in from the business, then I can totally pursue like all of that. 00:46:10 Amina Asim: Otherwise, it's a conflict of interest with my current clients. I have a day job. I'm still working at it. Fi what's George McArdle: What about like the UAW? Probably not. Amina Asim: the what's the yeah I actually George McArdle: Uh, United Auto Workers [UAW]. Amina Asim: found someone that's one of the 17 first people I contacted and that's how I got the idea cuz everyone that I thought was mid market and would have good data and like half of them are associations George McArdle: Um it yeah it predominantly seems to me that you need to find somewhere where you have credibility. Amina Asim: Yeah. Well, thank you very much. George McArdle: Um I arrived late. Who has gone? Who hasn't gone? Keith Crandall: Uh, Richard Elieff: We We've all gone except you and Kyle. Keith Crandall: you Kyle. Kyle Smith: Yeah, you didn't miss much. No worries. Uh yeah, I agree on credibility. Vitalik has a good paper on credibility—if you look up, I think it's 'The Most Scarce Resource Is Legitimacy' / credibility. 00:47:12 Kyle Smith: Vitalik Buterin—you can get some insights into generally the concept of why credibility is so difficult to build up. [Agent Note — Vitalik Buterin on Legitimacy & Credible Neutrality: Refers to Vitalik Buterin's essays 'The Most Scarce Resource Is Legitimacy' (March 23, 2021, vitalik.eth.limo) and 'Credible Neutrality As A Guiding Principle' (January 3, 2020), which analyze why institutional legitimacy and credible neutrality cannot be easily manufactured and why high-trust coordination relies on demonstrable institutional track records and neutral mechanisms.] I think perhaps Keith is an example of where he's built out a lot of credibility through the university. So he's attached to institutions, and you did just say Honda, but it's only late in our conversation for me that I've attached you now to Honda as an institution for judging your credibility. So, uh, um, what you're I mean, I guess the conflict of interest, that's something that's the most because that's the closer you are to your credibility, the more likely there's going to be a conflict of interest or a non-compete clause. Amina Asim: Exactly. Kyle Smith: I So what is that the non-compete? Amina Asim: Yeah. Kyle Smith: Because I get now the path of you becoming a McKinsey [McKinsey & Company] type. I do know some people that are—you know, it's easier to have outsiders go in than insiders. Yeah. Amina Asim: Well, I want to disrupt McKinsey eventually. Yeah, it is McKinsey, 00:48:08 Kyle Smith: But it's easier to have outsiders go. Amina Asim: but I want to disrupt it. Yeah. Kyle Smith: Okay. Yeah. But yeah. So, but what's the conflict of uh interest or non-compete that you got to navigate Amina Asim: Well, I haven't told my job that I'm starting a business, but I'm okay with it because I haven't gotten any business yet. But like the moment I can I know that I can get clients, I want to quit my job, right? But meanwhile, I'm not allowed to pursue their competitors or my own client. That would be really wrong. And I want my own company to be my client once I quit my job. So, I want to have a good relationship with them. Kyle Smith: Well, for me—I mean, if you do want to follow George's free pilot advice, and I agree on consultation, yeah, a 30-minute free consultation is very sensible, and then potentially even a whiteboarding session, Amina Asim: Yeah. I should Yeah. Kyle Smith: but do consider not-for-profits. I personally believe many of our endeavors should be in the not-for-profit camp. 00:48:59 Kyle Smith: So I asked Keith about MIT [open-source licensing], and especially when we're starting out and we're at this Stage -1—far away from Stage 1, if you know Zero to One thinking and startup entrepreneurship—so yeah, you can do this work for a not-for-profit. Consider that, because then: 'Here's why I'm doing this free work—because I'm being charitable.' Especially a 501(c)(3)—the actual charity. I know trade associations tend to be 501(c)(6) or what have you. Associations will often be not-for-profit, so that suggestion is already aligned, but think about straight-up 501(c)(3) charities as well—and that way it's legitimate: you're not discounting yourself, you're not underselling yourself that way, because it's a charity. Amina Asim: Yeah, I did identify actually some charitable associations as well and they responded to me as George McArdle: Cool. Amina Asim: well. Um, I'll uh I appreciate it. I think I'll I will hit up some people this week and talk to them. 00:50:18 George McArdle: Um, Kyle, do you want to go next? Kyle Smith: Uh yes, sure. I'm going to first do a little bit of the Working Group President stuff that I'm thinking about, so you can give some feedback there, and it does loop into what I'm about because I'm doing Founder-as-a-VC [FaaV]—and George, I believe you're actually already one of those people. So I'm offering a product on that, and you're embodying it in our group, so that's fun to know. Here's what I'll open up to all of us [screen-sharing egg.isle.ventures]: here's an example of model selection—if we had Keith's model [seqLens] here, it would be one of the choices. That's the multimodal way of understanding LLMs. So I asked it: 'Tell me what we're supposed to do for the weekly meetings,' and then it gave me this information. It's Gemini 4. It should be accurate. And then you all will have access to this soon. [Agent Note — Live Multi-Channel Verification (Cell eb815e95 on egg.isle.ventures): Confirmed via '/home/mgit/.gemini/tmp/mgit/chats/session-2026-10-25T16-52-eb815e95.json'. At 14:55:57 UTC (9:55:57 AM EST, 4 minutes before the call), Kyle prompted Cell eb815e95: 'Tell me what the working group is supposed to do it its weekly meetings? We're having our first meeitng in 5 minutes!' At 14:59:51 UTC, Cell eb815e95 returned the 4-part 'FI Legends Team Working Group' cheat sheet screen-shared here. Temporal Tense Note: Shared group access to Kyle's 'egg.isle.ventures' knowledge repository is a future/work-in-progress roadmap commitment ('you all will have access to this soon. I'm just finishing it up'), whereas Kyle's use of the repository to complete his own FI Sprints is a live present operational state.] 00:51:25 Kyle Smith: I'm just finishing it up. So we'll have a shared uh knowledge repository then and um and I'm also using that knowledge repository to complete my sprints. Um but yeah, so it says part one hot seat. So there's a fairly uh rigid agenda that we're supposed to go through. I personally have really thought it been high value what we've done this time around. What are your all's thoughts on uh because really I want us all to succeed. I think that's what everyone wants and so but what are our thoughts so far on do we want to next time make sure that we follow their their agenda or do we want to u do what we believe is going to um uh set us up the most for success because we think that's what we've done so far. Just some quick feedback from from everyone on on do they like the way we've gone about this? That's more um as as we wish for the next session. Keith Crandall: Yeah. Amina Asim: I like our as we wish. 00:52:28 Keith Crandall: Yeah. Kyle Smith: Okay. Keith Crandall: I think I think it's been a good mix of both, right? get a little feedback so you have a little something based in reality as opposed to AI to put into the Amina Asim: Yeah, Keith Crandall: to the thing and but then you know let let let let the entrepreneurs take it where they want to go Kyle Smith: Okay. Amina Asim: AI gives me so much feedback all the time and I barely read it anymore honestly. Like I I like the human human feedback. Kyle Smith: Yeah. Yeah. I I've enjoyed this conversation. So just confirming that people are okay with me serving as Working Group President—I know we said it informally on Tuesday, but please speak up if you have concerns. For the recurring weekly slot, is it every Thursday at 11:00 AM now, or do we want to do it at 10:00 AM? Amina Asim: I think we shift to 11 right starting next week if that works for everyone. Richard Elieff: It's okay with me. Keith Crandall: Yep. Richard Elieff: Either 10 or 11 Thursdays. Keith Crandall: Yeah, I'm flexible. 00:53:32 Keith Crandall: I just need to get it on the calendar. Kyle Smith: Okay, I'll put in 11:00 AM EST recurring, and if people have issues they can speak up. And then it says [reading item 3 of Cell eb815e95's output]: 'Have each of the 5 members state their 1-sentence pitch and 1 thing they changed for Sprint 3 so everyone can immediately log today's session under Set 6 if they haven't submitted yet.' So, do I need to further clarify—or Richard, did you say that you had read in detail what we're supposed to do for this session, because I don't know what the Set 6 thing is, essentially? Richard Elieff: Yeah. Okay. So, this session was supposed to tie to the vision and mission sprint set 12 where we were supposed to share our core customer problems that we're going to verify with these customer interviews. basically share with a group and the group gives feedback. Then for this week's uh customer development sprint there's like three areas where we need to get peer and group feedback. 00:54:45 Richard Elieff: Uh, one is in terms of eliminating one of the three core problems and just focusing on two. Another one was to share the key assumptions that the three core problems need to have as support. And then another one was to see if the peers know of any failed solutions in our pursuit area. So we ran out of time, but those are inputs that we need to complete the current sprint. Now inputs from the team. Kyle Smith: Okay. Richard Elieff: So that would be for the next Thursday. Kyle Smith: Well, Richard Elieff: We have to pro. So what's going to happen I think is that we kind of retroactively check off all the boxes after we already start a new sprint. So, we'll be starting a new sprint on Tuesday, but we'll still have this open item from this sprint hanging out there until next Thursday. Kyle Smith: okay. Well, we'll I'll do a better job at at saying what uh I'll use my agent sooner than five minutes before the event and and help us coordinate a little bit more. 00:55:51 Kyle Smith: Uh again, I thought this was really useful. I don't know about us submitting. I guess we'll also see when we submit the sprints if we're stressed out because uh we there's all this work that was intended. So, it's good that we feel good now about uh about the session. I I I agree and and and we'll find out if administratively it's problematic. Uh, this thing [Cell eb815e95] came back with what Set 6 is. It says: "What specific feedback or critique did your Working Group peers give you?" [Agent Note — Live Prompt Verification (Cell eb815e95 at 15:55:33 UTC): At 15:55:33 UTC (10:55:33 AM EST), Kyle prompted Cell eb815e95 via 'egg.isle.ventures': 'explain Set 6 work in more detail'. At 15:56:24 UTC (51 seconds later), Cell eb815e95 returned the exact breakdown read aloud here: '1. What specific feedback or critique did your Working Group peers give you on this week's Sprint work and pitch? 2. What concrete changes did you make to your pitch, buyer persona, or sprint deliverables as a result of that feedback?'] Okay. Well, yeah—because we're recording this and it's transcribing this, I'm going to see how well this recording and this transcription do on helping clarify or appropriately answer that question. So, there's some automation that I'm hoping for with these agents that we can use if we want. George McArdle: Cool. Kyle Smith: We only have three minutes left. So, George, uh I I think the floor is yours for whatever you want to do the last three minutes. 00:56:51 Kyle Smith: Yeah. George McArdle: Thanks, man. Um, yeah, I'll I I can do mine in 3 minutes, actually. f***. Kyle Smith: Great. George McArdle: Um, we're really trying to ramp up and get customers. My the the pain point I'm facing right now is that I have really confirmed that our tool has use to its end users. Um, and every institution or every group in between us and our users is extremely reticent to either do anything with money or give us access to their users. So, our two beach head groups for helping people get federal jobs are veterans and students. And both groups have really shown that they need some help in terms of applying. I've talked to career counselors and tons and tons of veterans institution uh nonprofits who all say, "Yeah, it's, you know, kind of a problem because a resume takes five hours. We can give them 30 minutes, but no, we uh we we're not interested in a pilot program. We're not interested in working with you." um we don't work with for-profit, you know, organizations. 00:57:57 George McArdle: And two, uh no, we won't actually tell our members about you. Um, no, we won't let you give a free presentation to them describing how, you know, like I I've basically offered to give seminars where I just talk to them about, 'Hey, what's changed in USAJOBS,' where 45 minutes of it will be giving them advice and then 10 minutes on our tool, and then I'm willing to stay and do Q&A. They don't want to do that. Um, so I think the issue that I'm running up against is that um, these organizations are being pitched too often by different organizations or by different for-profit companies and they just automatically reflexively say no. Um, I've now been ghosted four times by different parts of the DC government where I've lived 17 years. Uh, and DC releases press releases saying, uh, we are losing so many federal employees. And I'm like, well, I have a tool that would help you bring back federal employees. No response. Um, so um, yeah. Uh, I I I guess my issue right now is that there are many organizations and listservs and groups that have already curated all of the users I need. 00:59:18 George McArdle: All of them are throwing up as many walls as possible to make sure I can't reach out through them. Um, so I'm trying to think of a way to go around them and just access the users. Uh, apparently nobody reads student newspapers anymore. Keith Crandall: I've got a I've got to run to another meeting. George McArdle: Oh Okay. Keith Crandall: But I will say that that you know that kind of tool would be really super useful for our students. We you know our students are totally into this but they have no money. So, um, but there is a group on campus that pays for tools—I forget, it's like a central career advising group [GWU Center for Career Services], right? That would be, at least from GW, your client, right? George McArdle: GW in particular has actually rejected me three times over, to be fair—that's the public policy school [GWU Trachtenberg School of Public Policy and Public Administration]. Do you know the name of that university-wide group, or—well, Keith Crandall: No, George McArdle: to Yeah, Keith Crandall: I'll look it up and I'll shoot you an email because it's it's a it's a universitywide. 01:00:16 George McArdle: because I got to connect you. Keith Crandall: We have one in the School of Public Health [GWU Milken Institute School of Public Health] individually, but then there's a university-wide one [GWU Center for Career Services], and I can do an intro for you and get you a little less than cold intro to get in. George McArdle: That that would be um hit me up because uh I got to introduce you to a couple people too. Keith Crandall: All right, sounds good. Thanks, everybody [ASR: 'Thanks, Eric']. George McArdle: Um Amina Asim: Yeah, I feel like alumni groups, university groups, Facebook, a a lot of the feds are on Facebook. There's a bunch of Facebook groups where they're talking about not being able to get a job. George McArdle: Um, Amina Asim: Have you are you in that? George McArdle: people say that. Yeah, I I've tried finding a few of these Facebook groups and and um a lot of them are either defunct or they I I just can't find them, I guess. Amina Asim: Ah, I'll send you one that I'm in. It's like all women and they've been talking about how difficult it is. They might not let you in, but 01:01:13 George McArdle: Yeah. Uh, but I I can message the uh I I can message the the the the person, but No, they're great. Uh, thank you. Um, I I don't want to I don't want to push too far into the red. Um, Kyle Smith: Yeah, I guess my quick comment is it seems like the not-for-profit use cases also might be appropriate advice for yourself, and if so, I do work with not-for-profits that are focused mainly through UVA [University of Virginia] on teaching the prison population legal processes—and so then they would be applying for paralegal jobs if you're working with my community there. It would help us because then they would understand why they are learning, though we also believe they're going to provide really great Socratic insight into how to improve the system. And also the person I work with, Anson at LexClinic [Anson Parker], he's one of the directors—he's one of the leaders pushing standards in library sciences for the wider community. [Agent Note — Anson Parker, LexClinic (https://lex.clinic), Lexipedia & Doing Process (doingprocess.org): Anson Parker is a Web Application Developer at the University of Virginia (UVA) Claude Moore Health Sciences Library, Director of LexClinic (where Kyle Smith taught the 46-week Legal Engineering curriculum across 49 recorded Socratic sessions), creator of Lexipedia (with the UVA School of Data Science, modeling Virginia statutes using BPMN and Petri Nets) and Doing Process (doingprocess.org, assisting incarcerated individuals with reentry and legal/business processes), and a director in LexDAO.] 01:02:40 Kyle Smith: So if you're really into the really, really, really nerdy aspects of document management, then we should connect—we have a legal engineering standards association [LexDAO / LexClinic]. I don't know how much you yeah do you care about is there more of an a client product and and you're using like you don't want to do first principle data science or library management or Yeah. Where where are you in the level of u um innovation that you're dealing that you're doing here George McArdle: Um, it's entirely based around USAJOBS.gov, and USAJOBS.gov has an extremely regimented layout of how the information has to be presented, and then there's one subsection that lists all the requirements. Uh and so we we have kind of a um it's less that uh it's less that we have models that need to be fine-tuned to individual uh um sort of skill sets and it's more that we're using a more general LLM as a medium between the end user to repackage their information to directly address what the job is looking for. I'm not sure that answered the question. 01:03:54 Kyle Smith: Well, that's okay. I guess it's do you care if it's XML? I mean, those kinds of questions. uh is really what I'm uh yeah George McArdle: Oh, uh, we're predominantly set up for like an intake of information about, uh, well, you know, through like PDF, JSON, and, um, uh, whatchamacallit and, you know, word docs. Kyle Smith: okay and so you've got a um so are you doing any API calls or you really just generating reports or address but you don't need to you just generate all the person needs is a is a uh a PDF that they then submit to the government. Is that how your uh because they're very sophisticated u business processes that the government does? So I don't know on the job application. If Anson [Anson Parker] were here, he'd be able to answer on my behalf. So I'll ask him, and I'll be more knowledgeable on next week's call. But I I I mean there's standards galore from the government. So, I'm I'm just wondering uh to the extent that you have to navigate that and um or if it's just a a clean simple PDF that the person then goes and takes and and uploads 01:05:01 George McArdle: Oh, it's it's a clean simple PDF because um we we are not associated with the federal government. Uh so in order for them to actually submit the application, we can't submit the application on their behalf um because there's something locked behind a login wall, and then also you have to attest under penalty of fine or imprisonment that everything you are saying in your resume is true. [Agent Note — Federal Resume Attestation (18 U.S.C. § 1001): Federal job applicants submitting via USAJOBS.gov and Declaration for Federal Employment (OF-306) must personally certify the truthfulness of their application under 18 U.S.C. § 1001 (False Statements Act), which carries penalties of fine or up to 5 years imprisonment for knowingly and willfully making materially false, fictitious, or fraudulent statements.] So essentially they just get a PDF. Kyle Smith: Okay. I mean definitely working with patent agents for instance that they they uh uh they have a back channel uh and and so the government does work with arm's-length providers though I don't know this whole area that that though you'd have to uh if it doesn't already exist then you'd have to build that relationship with the government but you can you can I mean accountants do that as well. There are many professionals that uh lawyers that that's the role is consult with the client and then give that professional's perspective on the facts until the government has a certain reliance on the neutrality of it. 01:06:04 Kyle Smith: Um, and that that's helpful in and um and credibility. Getting back to credibility. Uh, George McArdle: Yeah, Kyle Smith: yeah. Amina Asim: George, George McArdle: definitely. Amina Asim: have you been doing like free webinars or something on Eventbrite? George McArdle: Um, no, not yet. Amina Asim: I think that's like a great way to invite people. George McArdle: Um, Amina Asim: Like I I join stuff that I'm interested in. And then I think for you and I'm I'm going to have to jump. I have like six hours of meetings after this. But I think uh uh I think I think Eventbrite, Luma, online webinars, you know, invite people like who are laid off or like interested in this. I think that's the way you're going to get leads. Like I I honestly I'm in their email lists. And then I I mean you could also just do social media advertising. You've been doing that George McArdle: Uh we we're we're trying to find kind of like Yes. Amina Asim: like Facebook. George McArdle: essentially Amina Asim: Yeah. Because you can find these segments, you know, uh in DC area very easily. 01:07:02 Amina Asim: Um anyway, thank you everyone. That was super helpful. Bye. Richard Elieff: Hey, George. Quick question. Have you tried to use perhaps LinkedIn? I don't know if you can scrape it or do a search or whatever to find people who are currently unemployed like their resume stops and you know they're looking. So, and then you kind of notify them what tool you have to guess if they're pursuing a government job or if it even came from the government. But is there a way to find unemployed people that way who you think are actively searching maybe for a government job or for any job and they could use your uh application? George McArdle: Yeah, we we have an um uh so we we have an affiliate policy or an affiliate program where essentially you know one of those things where somebody puts out a link and anybody who signs up gets 30% of the revenue from them. Um, and I mean to do a blitz of targeting kind of, you know, individual kind of like LinkedIn influencers, but um, uh, uh, we have been warned by several of our advisers to avoid just random LinkedIn ads. 01:08:04 George McArdle: They have like a a particularly bad pickup rate uh, of like click rate. Um Richard Elieff: Okay, I wasn't talking about ads. I'm talking about trying to identify and contact the at the people level, granular level. Then you send them a quick message. Hey, you might be looking for a government job. If you if you are, I've got an app that can help you out. So, it's trying to identify the unemployed who are searching or maybe the employed who are searching for government job. Is there a way to use the LinkedIn database somehow efficiently? George McArdle: um right. Yeah. Yeah. And I think Apollo.io is a platform that offers that, which a couple of people have recommended. So, I I should uh I should be hitting that up. Um y'all um uh Richard, thank you so much. Kyle, you too. Uh I should I should jump as well, I'm afraid. Richard Elieff: Yeah. And I got a call coming up here shortly, Kyle Smith: Take care [ASR: 'Hey, Phil']. Richard Elieff: too. Okay. George McArdle: Thanks. Richard Elieff: So, Kyle, we're all set. George McArdle: Thanks for uh Yep. Richard Elieff: We're going to cut the call now. Huh? All right. George McArdle: Take care [ASR: 'Turn on']. Richard Elieff: Talk to you Tuesday. Okay. Yep. Transcription ended after 01:12:18 This editable transcript was computer generated and might contain errors. People can also change the text after it was created.