The Proof That Doesn't Prove: Why ZK Inference's Single-Digit Milestone Has a Model-Size Caveat

Attestable emerged from stealth this week with a $20M seed from Altimeter and TLV Partners and a claim that sounds impossible: zero-knowledge proofs for LLM inference at single-digit overhead. Their demo proves inference of Meta's Muse Glimmer 30B on a single H100 at 85 tokens per second, generating compact, quantum-resistant proofs verifiable in sub-second time. Vitalik Buterin — who has been evangelising the overhead ratio as the metric that matters — noted this puts ZK proving overhead "approaching single…

Tech Briefing — Friday, August 14 2026

Grok 4.6 and Grok Bot headline a packed model day, with DeepSeek V4-Pro and Qwen3.8-Max also launching. Gemini crosses 1B users, OpenAI's COO departs, and Google shuffles DeepMind leadership. UK swelters at 38.1°C while settler violence escalates in the West Bank.

• 🤖 Models and Launches 6
• 🛠️ Agents and Tools 3
• 💰 Funding and Business 2
• 🏢 Industry Moves 2
• 🔒 Security 3
• 🌍 World and Australia 6
• 🔥 Hacker News 7

Reflective Essay

Today's field produced unusually consistent concrete anecdotes — almost every candidate found a specific tool-misuse story rather than defaulting to abstraction. The recurring surprise was how many essays landed their opening scene only to fumble the ending with a call-to-action or LinkedIn closer; the winner distinguished itself by letting the final line linger instead of resolving. Deepseek's 891-word reasoning trace for a 193-word essay is worth noting as an efficiency curiosity.