All articles
An OpenAI researcher "felt the AGI" watching AI agents work together. The real mechanism is smaller and pricier than the hype, and you can try a version of it in three browser tabs.
October 12, 2026 AI AgentsA rogue agent incident at Wikimedia and a sandbox escape Ethan Mollick described share one pattern. Here's the scope preamble that shrinks the gap in your next agent session.
October 08, 2026 AI ToolsA Utah startup's AI now sends acne prescriptions straight to pharmacies. The surprise isn't the tech: nobody has decided who is legally responsible when it gets one wrong.
October 06, 2026 AI ToolsAsking an AI "are you sure?" about a source just gets an apology. Asking it to reproduce the exact quoted sentence exposes fakes far faster. Here's the prompt and the two-step check.
October 01, 2026 AI ToolsLong AI chats drop your corrections and repeat old answers. The model didn't get dumber, it ran out of room. A two minute test and a restart habit that fixes it.
September 29, 2026 AI ToolsAsk a chat tool for a random number between 1 and 20 and 17 keeps winning. It isn't rolling dice, it's copying a human habit. Run the five minute test yourself.
September 25, 2026 AI ToolsTransformers can load llama.cpp GGUF files directly. Outside the narrow Metal fast path it dequantizes at load time, so you get the disk savings without the quantized kernels.
September 23, 2026 AI ToolsPrompt caching charges a premium to write and a discount to read, so it only pays past a break-even reuse count. Here is the formula, the usual bug, and a script that prints a verdict.
September 16, 2026 WorkflowsHuman defect detection collapses past roughly 400 diff lines, and agents blow through that in one prompt. Tier pull requests by blast radius with a deterministic CI gate instead.
September 10, 2026 AI AgentsAnthropic publishes the literal system prompt behind claude.ai. Ignore the transparency angle: it's a production-tested template for structuring your own agent's instructions.
September 03, 2026 AI AgentsThe usual advice for agent-ready data is add context, add metadata, add lineage. Tool-calling reliability improves faster when you delete the overlapping fields instead.
August 31, 2026 AI AgentsRunning agent-generated code in a plain Docker container trusts a boundary built for reviewed workloads. MicroVMs give you a real kernel per execution at roughly 125ms boot.
August 20, 2026 AI AgentsA test written by the same agent that wrote the implementation is a mirror, not a check. Which TDD invariants survive agent authorship, and the mutation-testing gate that catches the rest.
August 18, 2026 AI ToolsContext engineering borrowed cache vocabulary but skipped the part that matters: key design. Here's how versioned, provenance-based keys fix stale retrieval where TTL clocks can't.
August 17, 2026 AI AgentsOpenAI's Ultrafast tier hits 750 output tokens/sec on Cerebras hardware. Whether that changes your agent's wall-clock time depends on one ratio you can measure in 20 lines of Python.
August 14, 2026 AI AgentsMeta's Muse Glimmer is 30B params under Apache 2.0 and ships with no documented tool-call schema. Three concrete audit steps to run before it touches your agent stack.
August 10, 2026 AI AgentsPer-call RBAC can't say "deny this refund because a cancellation already happened." Sequence-aware authorization for AI agents is two Postgres tables and one plpgsql function.
August 10, 2026 AI AgentsRole confusion exploits how text sounds, not which role tag it carries. The fix isn't a filter — it's a trust-level envelope in your message construction code. Here's the structural patch.
June 24, 2026 AI AgentsPublic leaderboards score tool calling on clean synthetic schemas, not the nested mess your MCP server exposes. Here's the ~50-line Python harness that settles the debate on your own stack.
June 19, 2026 AI AgentsLong agent loops collapse around turn 8 — not because the window is full, but because tool results silently eat the budget. Here's the audit pattern and checkpoint prompt to fix it.
June 01, 2026