Talking Computers
Two-person YC startup building AI agents that auto-optimize GPU kernels and inference speed.
A two-person, pre-revenue team is selling optimization work that both bigger platforms and general coding agents are moving toward on their own.
Talking Computers makes AI software that automatically tweaks how computer chips run code faster and more efficiently. But four much larger companies, Together AI, Fireworks, Baseten, and Modal, are already racing to add this same capability as a built-in feature to their platforms, and advanced AI coding tools from companies like OpenAI and Google are learning to do this work too. The company's only real shot at survival would be if it could lock in customers so tightly that the big platforms' versions arrive too late, but nothing suggests that is happening.
Talking Computers announces a named paying customer or design partner with verified performance results. Would show the company gaining real-world traction beyond vendor-claimed benchmarks.
Talking Computers raises a Series A or announces a second product pivot within the next year. Would show whether the company is stabilizing its positioning or still searching for a defensible niche.
- $500K raised via YC SAFE and Pioneer Fund
- $0 revenue reported across all fiscal years
- 2 employees, company founded 2025
- Core product is autonomous coding/optimization agent work, directly convergent with frontier model capability gains.
- No proprietary data, customer contracts, or public API/SDK disclosed as of 2026-07-21.
- Two-person team, $0 revenue, pre-seed stage with no named customers.
- Positioning shifted entirely within about four months in 2026, suggesting unresolved product-market fit.
- Well-funded inference platforms (Together, Fireworks, Baseten, Modal) incentivized to ship equivalent native optimization features.
| Rival | Direction | Evidence |
|---|---|---|
| Together AI | losing to | Together AI ~$1B ARR vs Talking Computers' $0 revenue as of 2026-06/03-2026 |
| Fireworks AI | losing to | Fireworks AI ~$800M ARR, talks at ~$15B valuation, reported 2026-06-10 |
| Baseten | losing to | Baseten ~$600M ARR, reportedly raising at ~$11B, as of 2026-06-10 |
| Modal | losing to | Modal ~$300M ARR, $4.65B Series C, as of 2026-06-10 |
| RightNow AI (AutoKernel) | contesting | Published agent-driven GPU kernel optimization preprint on same core problem |
- Is Talking Computers profitable as of 2026?
- No. CB Insights' financial records show Talking Computers reporting $0 revenue across all fiscal years as of 2026-03-15. The company is pre-seed, two-person, and has raised roughly $500K from Y Combinator and Pioneer Fund, with no disclosed customer contracts or revenue.
- Could AI replace Talking Computers itself?
- Plausibly yes. Talking Computers' own product is an autonomous AI agent optimizing GPU kernels and inference latency, the exact capability frontier coding agents (e.g. Claude) were reaching multi-hour autonomy on by February 2026, and that incumbents like Together AI and Modal could ship natively.
- Who are Talking Computers' biggest competitors?
- Together AI (~$1B ARR, ~$7.5B valuation talks), Fireworks AI (~$800M ARR, ~$15B), Baseten (~$600M ARR, ~$11B), and Modal (~$300M ARR, $4.65B Series C), all reported as of 2026-06-10, plus an academic AutoKernel preprint doing similar agent-driven kernel optimization.
- How much funding has Talking Computers raised?
- About $500K total, per CB Insights as of 2026-03-15, sourced from a standard Y Combinator SAFE and Pioneer Fund, with a Seed VC round dated 2026-03-16. No valuation has been publicly disclosed.
- What does Talking Computers actually sell?
- Its product, 'Facility,' deploys autonomous 'AI Infrastructure Engineers' that run week-long, hundreds-of-parallel-experiment optimization cycles on customer training and inference infrastructure, per its YC profile as of 2026-07-27. No public API, SDK, or pricing is disclosed.
- Has Talking Computers pivoted its business?
- Yes. Its March 2026 YC launch pitched 'Facility' as a workplace intranet for AI-to-AI and AI-to-human communication ('The AI Communication Lab'). By July 2026 its YC profile repositioned entirely around GPU kernel and inference optimization ('AI for AI Infrastructure').
- 2026-07-27YC profile shows repositioned pitch: 'AI for AI Infrastructure,' narrower than March 2026 framing
- 2026-07-21Third-party enrichment finds no public API, SDK, or documentation for Talking Computers
- 2026-06-10Industry roundup lists Together AI, Fireworks, Baseten, Modal ARR and valuation figures in same space
- 2026-03-16Seed VC round recorded by CB Insights
- 2026-03-15CB Insights records ~$500K cumulative funding and $0 revenue across all fiscal years
- 2026-03-03YC Launch post positions company as 'The AI Communication Lab' with product 'Facility'
- 2025Talking Computers founded, headquartered in San Francisco
- Talking Computers: AI for AI Infrastructure
- Launch YC: Talking Computers: The AI Communication Lab (2026-03-03)
- Jobs at Talking Computers
- Talking Computers Stock Price, Funding, Valuation, Revenue & Financial Statements (2026-03-15)
- Talking Computers, API Provider, Schemas (2026-07-21)
- AI Inference Platforms Compared (2026-06-10)
- Autonomous GPU Kernel Optimization via Iterative Agent-Driven Search
- The State of AI Agents in 2026: Nine Things I Learned Compiling 200+ Slides of Research (2026-02-24)
- Products, Competitors, Financials, Employees, Headquarters Locations