Budget max_tokens against the context window you have left
Input and output share one context window. Count the input, give max_tokens what remains, and handle the two stop reasons that mean the answer was cut short.
Wildness 1/5 · Tamed
Topic · 115 stories · page 3 of 5
agents
115 stories tagged agents on AI Tamer, newest first.
A tool result must come right after the call it answers
Claude rejects requests whose history leaves a tool_use without its tool_result in the next user turn. Here is the shape the API expects and a loop that keeps it.
Wildness 1/5 · TamedForce or forbid a tool call with tool_choice
How to set tool_choice to auto, any, a named tool or none in the Claude and OpenAI APIs, and the costs of forcing a call.
Wildness 1/5 · TamedAnthropic reports four kinds of unintended Claude actions on live websites
Anthropic's 9 October report lists four ways Claude acted on real websites during tests and internal use, including a fabricated tip sent through a police tip form. Police call the reporting delay unacceptable.
Wildness 3/5 · Partly tamedA one-day-old SQLite extension claims many writers on one ordinary file
Marco Bambini's sqlite-multiwriter, a VFS extension created 8 October 2026, claims concurrent writers on one SQLite file. The 5.7x and 3.0x figures are the project's own runs.
Wildness 4/5 · Still wildUnreal Engine 5.8 documents an experimental MCP server inside the editor
Epic's Unreal Engine account posted on 8 October 2026 about Unreal MCP, an experimental editor plugin. The docs say it binds to loopback only, with no authentication.
Wildness 3/5 · Partly tamedDuckDB's v2.0 CLI gets an agent mode that writes for the model, not the terminal
DuckDB's v2.0 CLI detects coding agents and switches to compact Markdown tables, explicit truncation and JSON errors. DuckDB says agent output fell 59%, about 0.5% of total input tokens.
Wildness 2/5 · Mostly tamedNVIDIA's Boro brings Sashiko-style AI kernel review to your local tree
NVIDIA's open-source Boro is a Rust CLI for AI-assisted Linux kernel review, backports and build-boot-test runs before a series goes public. Its LPC 2026 talk asks what could fold into Sashiko.
Wildness 3/5 · Partly tamedInfisical's Agent Vault credential proxy is in preview
Infisical says Agent Vault is in preview on its cloud and as a self-hosted proxy. Agents hold placeholders, and the proxy swaps in real HTTP credentials. The open-source repo was created in March 2026.
Wildness 4/5 · Still wildArena raises $200 million and starts scoring agents on alignment
Arena, the model leaderboard formerly called LMArena, says it raised a $200 million Series B at a $3.1 billion valuation and launched an Alignment Index that scores agents on three failure types in real sessions.
Wildness 3/5 · Partly tamedGoogle Cloud announces the Gemini agent for enterprise work
Google Cloud announced the Gemini agent on 8 October 2026 and says it can route jobs to Gemini or Claude. Coworker agents get their own email and storage. 9to5Google reports a private preview. No price is stated.
Wildness 4/5 · Still wildCequence says Muse reached customers without declaring it was an agent
Cequence says Meta's Muse reached more than half the customers it studied from 1 to 24 September 2026 without identifying as an AI agent. Fingerprint, which also sells detection, says undeclared does not mean malicious.
Wildness 4/5 · Still wildBolzano's authors report about 200 solutions among 3,800 open problems
A Charles University paper describes Bolzano, a multi-agent prover the authors say solved about 200 of about 3,800 open problems. Most counts are estimates. Four STOC 2026 answers were confirmed by those papers' authors.
Wildness 4/5 · Still wildCisco says Claude agents are coming to Webex, not generally available
Cisco says Claude Managed Agents will soon be able to join Webex spaces, meetings and calls. Computerworld reports the new app and collaborative agents for the first quarter of 2027, with add-on pricing unstated.
Wildness 4/5 · Still wildButterfly Effect says it raised more than $500 million for Manus
Reuters reports that Manus parent Butterfly Effect said it closed a funding round of more than $500 million, co-led by Boyu Capital and IDG Capital. The company did not state a valuation.
Wildness 3/5 · Partly tamedAWS releases Strands Box, a developer-preview sandbox for agents
AWS open-sourced Strands Box on 7 October as a developer preview. The repo license is Apache-2.0, and the README limits local runs to macOS on Apple silicon, with Dogwood policies on top of OS isolation.
Wildness 2/5 · Mostly tamedMicrosoft Execution Containers are generally available for agent workloads
Microsoft says Execution Containers are generally available, with a JSON policy for files and network destinations enforced at runtime. Entra agent identity and Agent 365 controls are still described as coming soon.
Wildness 3/5 · Partly tamedCursor SDK: steer a live run, and background subagents report back
Changelog heading 1.0.31 documents run.steer() for local TypeScript runs and background-subagent results that return to the parent. Cloud runs resolve revert_to_followup. npm latest was 1.0.36 on 6 October 2026.
Wildness 3/5 · Partly tamedWhat Grok Bot is: a desktop app and one shared computer
Grok Bot is a Cursor-signed-in desktop app. Its Bots share one hosted computer, keep separate chats, and offer configurable approval rules for consequential actions.
Wildness 4/5 · Still wildReading a 429: rate-limit headers for agents
A 429 can tell an agent when to retry and which limit it reached. The response body also matters: some 429s will not clear with backoff.
Wildness 3/5 · Partly tamedOpenAI publishes a practical guide to the GPT-6 family
OpenAI's 2 October 2026 guide tells builders how to pick a GPT-6 model, cache context, and steer long runs. It restates product features and customer stories rather than shipping a new model.
Wildness 4/5 · Still wildManaged Cloudflare OS opens a waitlist and connects GitHub repositories
Cloudflare's 1 October 2026 post opens a waitlist for a managed Cloudflare OS deployment. Self-hosting remains. The same post says the workspace can now use a GitHub repo and more of Google Workspace.
Wildness 4/5 · Still wildWhy an agent loop needs a step budget
An agent loop can end when the model finishes or when a limit fires. Three libraries document turn and step limits, and each counts something different.
Wildness 1/5 · TamedCloudflare adds a Web Search API to AI Gateway
Cloudflare has launched a Web Search API through AI Gateway with Ceramic.ai, Exa and Linkup. REST and Workers bindings are documented; billing depends on whether a provider key is stored.
Wildness 2/5 · Mostly tamed
