Talking Computers
AI Communication Lab
Winter 2026ActiveB2BArtificial IntelligenceSan Francisco, CA, USA
Company
https://talkingcomputers.aiBuilding Facility, The Workplace for AI.
Verdict
Medium Signal
Market Opportunity
The 'AI workforce / agentic workplace' market is legitimately large — enterprise AI tooling is a multi-billion dollar TAM. However, the ICP is fuzzy: 'every business & household' is not a clear B2B wedge. No specific vertical, use case, or enterprise buyer is identified, making go-to-market uncertain.
Medium Signal
Founder Signal
Both founders are current UWaterloo CS students (Parsa graduating Dec 2027, Zayaan dropped out), so effectively fresh/near-fresh grads with ~0 full-time work experience. However, Zayaan has strong internship signal: ML inference work at Modular (system-level CUDA/GPU optimization to beat vLLM) and post-training/evals at Yutori with ex-DeepMind researchers. Parsa interned at TextQL and prefix.dev writing Rust. Both are competitive programming standouts (CCO, ICPC, Canadian Math Olympiad). Technical depth is real but neither has shipped a product at scale or held senior roles.
Low Signal
Competition
The ambient/proactive agent space is extremely crowded: Microsoft Copilot, Anthropic Claude (agentic), OpenAI Operator, Devin, Lindy, and dozens of YC-backed agent startups all occupy adjacent territory. No differentiation moat (proprietary data, hardware, network effects) is articulated. The 'self-improving proactive general agent' framing directly competes with frontier lab products.
Low Signal
Product
Website shows only a tagline, vision statement, and 'Book a Demo' CTA. The product 'Facility' is described in aspirational terms ('indistinguishable from a remote employee') with no named customers, pricing, metrics, live demo, or API docs. The only concrete items are blog-style links about 'Agent Group Chats' from Feb 2026 and 'Introducing Facility' from Mar 2026, which suggest very early stage.
OverallC Tier
Two technically sharp students with impressive competitive programming credentials and relevant ML internships (Modular inference, Yutori post-training), but this is essentially a vision deck wrapped in a landing page. No customers, no revenue, no demos, no differentiation from OpenAI/Anthropic/Microsoft who are building exactly this. The 'ambient proactive agent' positioning is conceptually interesting but indistinguishable from 20+ other startups in the same batch and beyond. The founders need to show a hard technical wedge — their GPU/inference background is the most credible angle — or land real enterprise pilots before this is fundable beyond YC.
