YC Winter 2027 research brief: application facts, RFS, funded landscape and patterns (as of 19 Oct 2026)

Short answer: the next batch is Winter 2027. The on-time deadline is Monday 2 November 2026 at 8pm PT, which is 04:00 UK on Tuesday 3 November. YC asked for a "Company Brain" and an "AI Operating System for Companies" in its Summer 2026 requests. It has already funded at least five company-brain startups. It has also funded one direct "AI-native McKinsey" (Foaster, Spring 2026), and that is the closest competitor to this team.

1. YC application facts

Dates and terms

What this means for the team: Marton's multi-agent OS was built in Claude Code and Codex sessions, so a Paxel report is evidence YC can check. It is the strongest optional signal available to them.

Founder video: YC's own spec (ycombinator.com/video)

• "The video should be 1 minute long and should contain nothing except the founders talking."
• Every founder should appear. If you can't be in the same room, "screen record a video call instead."
• Content: "introduce yourselves, explain what you're doing and why", plus anything else about the founders or the project.
• "This is not the place to submit a demo or promotional video." Demos have their own field.
• "Do not recite a written script: Use bullet points instead."
• YC says it is statistically "much more likely to interview people who submit a video" (howtoapply).

Demo video spec

UNVERIFIED. I could not find an official length or file-size limit. The "3 min / 100MB" figure appears only on a vendor blog (Flowjam). Check the live form.

Form questions (verbatim, but from a third-party copy)

These come from a copy of a May 2025 application (Tnkr.ai YC application). The live W27 form needs a YC login, so the current wording is UNVERIFIED. - "Describe what your company does in 50 characters or less." - "What is your company going to make? Please describe your product and what it does or will do." - "How long have the founders known one another and how did you meet? Have any of the founders not met in person?" - "Who writes code, or does other technical work on your product? Was any of it done by a non-founder? Please explain." - "How far along are you?" / "How long have each of you been working on this? How much of that has been full-time? Please explain." - "Are people using your product?" / "How many active users or customers do you have? How many are paying? Who is paying you the most, and how much do they pay you?" - "Where does your revenue come from? If your revenue comes from multiple sources (ex. multiple products, multiple companies or a mix of consulting and this product), please break down how much is coming from each source?" This one matters here: revenue from coaching or consultancy work has to be broken out separately. - "Why did you pick this idea to work on? Do you have domain expertise in this area? How do you know people need what you're making?" - "Who are your competitors? What do you understand about your business that they don't?" - "How do or will you make money? How much could you make?" - "If you had any other ideas you considered applying with, please list them..." - "Where do you live now, and where would the company be based after YC?" / "Explain your decision regarding location."

"What have you built": older YC text asks for "something impressive that each founder has built or achieved" (howtoapply). Applicant copies from 2021 to 2023 say "the most impressive thing other than this startup" (Hard Pivot W21). The current wording is UNVERIFIED.

"Why now": this is not a standalone question in the 2025 copy. The nearest is "Why did you pick this idea…". Dalton Caldwell's talk says the form makes you work through why-now anyway (secondary: videohighlight summary).

What partners look for (concrete criteria)

• Clarity: the first interview question is usually "What is your company working on?" Answer in "a few simple, jargon-free sentences" (/interviews). Jessica Livingston: "Don't use buzzwords. Speak in plain language" (YC blog).
• Speed: "Demonstrating you can move fast and make quick progress is the most surefire way to impress your interviewers." Ship between applying and interviewing (/interviews).
• Knowing your users and metrics: know where users come from, retention and unit economics. YC "may ask for verification" of any numbers you give (/interviews).
• Knowing competitors: be "intimately familiar with the existing products in your market, and what, specifically, is wrong with them" (/interviews).
• Candour about obstacles: partners are "more convinced by candid discussion of difficulties than glib dismissal" (/interviews). PG's essay says the same: disclose flaws (howtoapply).
• Technical co-founder: "For most businesses, that usually means you need a technical co-founder." The team should build the product itself, not outsource it (/faq).
• Founder qualities (PG): determination, flexibility, imagination, naughtiness, and friendship: "They must genuinely like one another, and work well together" (paulgraham.com/founders).
• How long the founders have known each other: a form question. Seibel lists it among the first things reviewers check (secondary: mogulmillennial). The "at least 6 months" rule seen in guides is UNVERIFIED.
• Interview prep: "Don't rehearse"; every founder should answer at least one question (/interviews).
• Not found: standalone 2025-26 essays on application criteria by Harj Taggar, Jared Friedman or Gustaf Alströmer. Their views here come from RFS text only.

2. YC Requests for Startups: items relevant to this idea

I extracted these from the JavaScript bundle behind YC's own RFS page (ycombinator.com/rfs), which holds every season from Summer 2024 to Fall 2026. The current list is Fall 2026 (13 items). There is no Winter 2027 list yet.

No RFS uses the phrase "AI employees" verbatim. Items 1, 2, 6 and 10 above are the closest.

3. Funded landscape

How I built this: batches, one-liners and locations come from YC's public company directory, queried through the search key embedded in ycombinator.com/companies. Each company page is ycombinator.com/companies/<slug>. The "buyer" column is my inference from each company's own description, not a YC field.

Company brain and shared agent memory

Agent orchestration, company OS and "AI employees"

AI-native consultancy, transformation and agencies

B2B lead gen and GTM

I found no YC company aimed at proposals or sales ops specifically for consultancies. Treat that as "not found", not as proof there are none.

AI for product management

Funded non-YC competitors

4. Patterns: what YC actually funds in this space

1. YC funds the company brain now, which also means the generic version is crowded. Blomfield and Hu wrote RFS items in Summer 2026. At least five funded companies call themselves a "company brain" or a "shared brain" (Hyper, Memory Store, Savant, Hyperspell, Alloovium), plus close neighbours (Glen, Wato, Nessie). A horizontal "brain for everyone" pitch would be the sixth. Alloovium shows the alternative: a vertical company brain for one industry.

2. "Sell the work, not the tool" has been a consistent YC request for 18 months, and it has grown. The RFS line runs from Full-stack AI (Friedman, Summer 2025) to Government Consulting (Alströmer, Fall 2025), AI-Native Agencies (Epstein, Spring 2026) and AI-Native Service Companies (Alströmer, Summer 2026). My keyword count of directory one-liners ("AI-native firm/agency/brokerage", "full-stack AI") finds 6 in W26, 6 in Spring 26, 17 in Summer 26 and 4 so far in F26, a jump right after the Summer 2026 RFS. That count is my own heuristic, not a YC figure.

3. The funded services companies replace a named, outsourced budget. The examples are accounting (Balance, Last Accounting Company, Rational "First Zero-human Accounting Firm"), law (Moritz, LegalOS, Vector Legal), insurance brokerage (Panta, Covera), audit (Arden) and recruiting (Prism). Alströmer's reasoning: services spend is "many times larger" than software spend, and outsourced services are "much easier to replace."

4. The consulting wedge YC funds is the discovery step. Foaster, Ontora and Sona8 all use agents to interview employees and map how work gets done. Foaster promises a roadmap "in just 2 weeks". This is the closest overlap with a 16-year innovation consultancy. Expect partners to ask: "how are you different from Foaster?"

5. Partners want a measurable outcome, not a platform claim. Examples: Hu's "cut sprint time in half and ship twice as much", Foaster's 2-week roadmap, and the interview guide's emphasis on metrics that "we may ask for verification of." "B2B AI-native workflow orchestration" style copy is flagged as saying nothing (Causo).

6. Founders who lived the problem, with a technical builder on the team. The FAQ expects the founders to build the product. The form asks who writes code, and Paxel now scores real coding sessions. Kestral's founders led AI at Asana; Glen ingests the same Claude Code and Codex sessions its builders use. A working multi-agent OS the founder runs his own work on is exactly this signal, and the Paxel report is how to evidence it.

7. Tiny teams and revenue per employee. Many companies in these categories list 1-3 people: Pentagon 1, Glen 1, The Company Company 1, Foaster 2, Hyper 2, Savant 2. That fits the Fall 2025 "10-person, $100B company" RFS.

8. London is viable, but the batch is in SF. Directory location counts for London: 8 (W26), 7 (Spring 26), 3 (Summer 26), 2 (F26 so far). Examples include Acyclic, Trace, Prism and Hessian. The FAQ requires in-person attendance in San Francisco, and the form asks where the company will be based after YC.

9. There is an open lane next to the product-coaching background. Miklas's Spring 2026 "Cursor for Product Managers" RFS has only a few funded responses so far (Kestral, Scott AI, LightSprint, OpenSpec). All are engineering-team tools; I found none built by product coaches.

Still UNVERIFIED (check on the live form before writing answers): - The exact current W27 form wording. - The demo-video limits. - Whether "Why now" is its own question.

Nothing was written to disk except temporary scratch files.