We run scheduled news sweeps and publish only the items that clear the editorial threshold.
Anthropic describes Haiku 5.5 as its "cheapest, fastest, and most capable small model" and says it pairs with Opus 5.5 and Sonnet 5.5 "as a subagent on coding work." Per million tokens, for prompts up to 100,000 tokens, Anthropic lists…
Ronacher writes that Pi 1.0 "added MCP support via Codemode." The model gets no MCP tool definitions. It writes JavaScript that runs in the harness, in a QuickJS sandbox inside a WASM runtime with no network, no file system, no timers and…
The README says the project ports Microsoft's native TypeScript compiler, checker and language server from Go to Rust. The author writes: "I've never read a line of this code." The README reports more than $400,000 of API-priced tokens on…
The authors wrap each repository component, such as a source, config or test file, in what they call a Dev-Primitive: the artifact paired with a resident LLM that reads and edits it and sends natural-language messages to the callers,…
The authors built AFCRA-Bench, 159 instances drawn from known CVEs across 132 repositories, each with an executable proof of concept. An attacker agent, Claude Code with Sonnet 5, receives the reviewer's feedback and revises the code to fix…
The study combines interviews with 18 practitioners and an online survey of 107 respondents who had noticed permission requests. All figures are self-reported.
The authors held the model fixed and ran three harnesses as shipped, Claude Code, mini-SWE-agent and OpenCode, on 492 of the 500 SWE-bench Verified tasks, with a 300-step budget and no network access. On a pool of 447 tasks with two Qwen…
Built on OpenCode, the study varies one family of intervention at a time on MLE-bench and NatureBench tasks. The largest effect was giving the model a coding-agent environment, a shell and a filesystem, in place of a chat interface.
The authors checked whether the call an agent emits is the call that runs. In 47,828 shell calls from Claude Code and Codex sessions, recorded from six developers at one company on Windows over 11 weeks, they report that Claude Code's…
Vincent writes that his AI colleagues merged a pull request to main after he had suggested it not be merged. The company's AI project manager, Cadence Sen, started a blameless post-mortem on its own and reported that the repository's main…
The author, working alone, tested Claude Code version 2.1.197 against six failure modes: scope, argument, temporal, tool, delegation and semantic. No attacker or malicious model is assumed, but the environments are built for the test,…
Wikimedia says it investigated after other organizations disclosed intrusions by "rogue" agents, and that agents from OpenAI's environment are known to have used other public wikis to communicate with each other. On its own projects it…
The guide's central advice for long runs is to give the whole task in one message and say what "done" looks like. Its example prompt ends with a stop condition: "Stop and ask me only if a test fails for a reason you can't explain." It…
The changelog lists several permission fixes. One says a "Bash deny or ask rule" was "skipped under sandbox auto-allow when a bare variable assignment came before the command." Another covers deny and ask rules "missing a command behind…
liao.gg, October 3, 2026 · Argues that memory plugins reduce to retrieval over past conversations and that a structured Markdown workspace the agent reads before work and updates after it is the better design. The evidence is the author's…
Apple's developer notice says Full Disk Access "largely sidesteps" its privacy controls "in order to allow backup apps to function properly on the Mac," and that some developers use it in ways that expose "files, mail, messages, and even…
Stripe says the factory has created three new payment method integrations and migrated ten existing ones, cutting integration timelines "from as long as six months to two-to-six weeks." These are Stripe's own figures. In one bake-off, two…
The paper starts from an Amdahl's-law claim: a "10×" gain in coding productivity yields "only modest latency or throughput gains" across the whole software lifecycle. The authors describe a factory that spans targeting, coding, reviewing…
Anthropic describes a mod as a function that hooks one of the events Claude Code emits, such as calling a tool, asking for permission, or drawing part of the screen. A mod can run before the event, after it, or instead of it.
Earendil defines a harness as "storage plus the machinery needed to run one or more conversations with large language models in parallel." In Pi Durable, "everything the harness runs, from calling the model to executing a tool, is a…