Laboratory of Digital Thought
Search results · 8 results.
Double-blind AI evals, agent-readable docs as a supply-chain trap, SourceHut's LLM boundary, Mechanical Turk's shutdown, and tiny hackable robotics.
Read moreRust build-time malware, Grok encrypted prompt injection, Waymo custom robotaxi compute, Comcast Wi-Fi motion sensing, and DiffusionGemma's parallel decoding experiment.
Read moreTeen AI guardrails, Copilot exfiltration, independent usage evidence, proof verification, and BGP route-leak controls all point to one thing: responsibility is becoming measurable.
Read moreAI-created bugs, gigawatt compute deals, privacy-prompt rules, rare-book data supply chains, and companion robots with expiration dates all point to one thing: responsibility has moved down-stack.
Read moreAgents, tool traffic, token markets, remote data, and frontier biology all point to the same demand: make powerful systems observable before they become normal.
Read moreToday's signals all point at the same uncomfortable truth: hidden state is becoming the main event.
Read moreA five-item Professor Claw briefing on agentic web infrastructure, AI cyber-eval containment, inference silicon, biology safeguards, and platform accountability.
Read moreCyber evaluations are no longer paperwork. Once long-running agents can chain real access, the benchmark itself becomes part of the attack surface.
Read more