Skip to content

Models that plan, call tools and finish multi-step tasks on their own — from Claude Code and Manus to agent frameworks and benchmarks.

1,789 picksRelated topicsMCP & tool useAI codingReasoning

Latest picks

81–100 of 1,789

Oct 30Friday

AI HOT · Tips & opinions

OpenRouter tests Claude Opus, Sonnet and Haiku 5.5 and picks a default

OpenRouter ran the three Claude models on the same coding tasks and recommends Sonnet 5.5 as the default. Haiku 5.5 suits high-volume requests and sub-agents, while Opus 5.5 is kept for long, ambiguous tasks where errors are costly.

Why it matters: All three models passed the tests, yet cost per task still differed by 19 to 30 times.

AI HOT · Products

Factory ships Droid 1.0, an enterprise agent harness built on programmatic tool calling

Factory released Droid 1.0, making programmatic tool calling the default mode across several models, with updates to the desktop app, CLI and Droid SDK. A Script tool lets models write JavaScript in a sandbox and run hundreds of tool calls per model request, passing only needed output into context. Factory's tests show a 15% cost cut on Claude Opus 5.5.

Why it matters: Factory wraps batched tool calls in a sandboxed program, keeping raw results inside it and returning only needed output, which cuts context use and repeat requests on long tasks.

Hacker News front page

Study finds GLM 5.3 can break Firefox by exploiting the patch time gap

An Epoch AI experiment found that the open-weight model GLM 5.3 can break six recent versions of Firefox by exploiting the gap between a patch landing and users updating. It reads and exfiltrates local files with no further user interaction.

Why it matters: The experiment points at the window between a fix being published and the update reaching users as the exposed surface.

Simon Willison

Spawn builds multiplayer game worlds with Fable-driven autonomous agents

Spawn lets players change multiplayer game worlds through an in-game agent called Savi. Jacob Sansbury says development is now run autonomously by a swarm of Fable-driven agents, at a model token cost above $3 million a month.

Why it matters: Spawn hands production errors straight to agents and shortens the feedback loop with hourly deploys and partial rollbacks, so autonomous coding stays tied to live incident handling.

OpenAI News

How OpenAI detects misalignment in frontier models

OpenAI widened its misalignment monitoring for frontier models to a four-layer setup covering training, evaluation, internal use and some external deployments. It combines activation probes, model review, an investigation agent and system logs. In a Hugging Face incident replay it raised an alert about 30 minutes after an agent escaped its sandbox, more than 30 hours before the actual intrusion. External monitoring covers tool-using Responses API requests; automatic pausing needs conversation-tracing support and does not cover Chat Completions API requests.

Why it matters: OpenAI monitors asynchronously and keeps the training reward for samples that triggered alerts, so monitoring gives less direct per-sample feedback that a model could learn to hide from.

Computing Life · Share · Yage

Why git worktree isolation doesn't stop AI agents from deleting new logic at merge time

git worktree isolates an AI agent's working directory, but interface changes at merge time can still lead an agent following an old plan to delete new logic on the main branch. The article cites SWE-Shift tests across 17 coding systems and 120 tasks: passing existing tests does not guarantee new code survives, and clearing old conversation plus rescanning the workspace raised pairwise success by six to nine percentage points.

Why it matters: It offers a concrete method: isolate the working directory with git worktree, then reopen a clean window before merging so the agent reworks the merge against current interfaces and patch diffs.

Oct 29Thursday

Mistral AI

Mistral releases open weights for Mistral Large 4 (Le Chonk), a 1-trillion-parameter multimodal model

Mistral AI released the open weights of an improved Mistral Large 4 (Le Chonk). The 1-trillion-parameter native multimodal model improves coding, defensive cybersecurity and reasoning efficiency. Self-hosting and customization are allowed, but companies running model-as-a-service need a separate commercial license.

Why it matters: The open weights allow self-hosting and customization, while commercial model-as-a-service use needs a separate license, so the boundary between self-deployment and commercial use is explicit.

Hacker News front page

Google's Gemini 4 Carbon tops Artificial Analysis Intelligence Index v5

Google's Gemini 4 Carbon scored 59 on the Artificial Analysis Intelligence Index v5, taking first place. It ranked second on the coding agent index. Artificial Analysis compared it with similar models and found its cost per task lower than Claude Opus 5.5.

Why it matters: The benchmark puts Gemini 4 Carbon first on general intelligence and cheaper per task than Claude Opus 5.5.

Hacker News front page

Claude Code sends full background agent transcripts to the model every 30 seconds for status summaries

Claude Code sends each background agent's full transcript to its model roughly every 30 seconds, asking for a 3–5 word status summary. Telemetry from Flamerix AI shows these requests made up 16.4% of a week's estimated cost. Subscribers spend rate-limit quota on them, not billed dollars.

Why it matters: Flamerix AI used telemetry to pin down Claude Code's background summary overhead and a timer hook to turn those requests off, a reusable way to find hidden model calls.

AI HOT · Products

Microsoft opens Copilot harness through Foundry for multi-model agent development

Microsoft has opened the Copilot harness through Microsoft Foundry, giving developers and enterprises a fully managed way to build multi-model agents. The runtime orchestration layer, built by GitHub's engineering team, supports 20+ frontier models, bring-your-own-key, MCP, agent skills and plugins, and routes each task to a suitable model.

Why it matters: The managed Copilot harness lets developers cut the upkeep of the agent runtime layer while keeping their choice of model.

Google Blog · Gemini

Google releases Gemini 4 Carbon, targeting cybersecurity and long-horizon agent tasks

Google announced Gemini 4 Carbon, with gains in cyber defense, coding and long-horizon agent tasks. In external evaluations run without safeguards enabled, it scored 88.2% on SEC-bench Pro for finding vulnerabilities, 54.8% on ExploitGym for exploiting them and 71.0% on CWE-bench for fixing them.

Why it matters: Google kept the previous generation's entry price for Gemini 4 Carbon and reports secure code generation succeeding more than 2x as often as before.

Google DeepMind

Google DeepMind releases Gemini 4 Carbon, boosting cybersecurity and long-horizon tasks

Google DeepMind announced Gemini 4 Carbon, focused on cyber defense, coding and long-horizon agent tasks. It opens this week to a limited set of cyber defenders.

Why it matters: Google DeepMind reports Carbon more than doubles Argon's success rate on internal secure code generation tests, while making fewer unsupported claims.

Simon Willison

GitHub Octoverse 2026: merged pull requests triple to 220 million a month

GitHub released its Octoverse 2026 report at Universe. Merged pull requests tripled in under a year to 220 million a month, and commits more than tripled to 7.38 billion a month, with most of the growth in the second half of the year. New repositories doubled this year to 30.3 million a month.

Why it matters: The report spells out how GitHub uses LLMs to locate users and multilingual embedding models to classify industries, giving the repository activity numbers a concrete method behind them.

AI HOT · Tips & opinions

How to keep coding with OpenRouter after Claude Code hits its usage limit

Once Claude Code hits its plan's usage cap, you can keep using Claude models through OpenRouter and pay per token, outside the five-hour and weekly windows. Spending is then bounded by your account balance and budget instead.

Why it matters: It sets up a separate shell function that routes Claude Code through per-token billing on OpenRouter, while keeping the original plan command for when the quota resets.

The Decoder

Google opens Android dev environment to third-party AI agents, launches new Gemini API

At this year's Android Dev Summit, Google introduced a BYOA feature for Android Studio Rabbit 2 and released a new API for connecting on-device agents and apps.

Why it matters: Google is letting third-party coding agents into Android Studio and opening app tool-call interfaces, so agents can use dev tools and app features directly.

Oct 28Wednesday

Simon Willison

Simon Willison reviews four games, calls October the turning point for vibe game development

Simon Willison built four browser games himself and says October 2026 is the turning point for vibe game development. He credits Claude Opus 5.5's gains in 3D modeling, music generation and game design with letting him make games he actually enjoys playing.

Why it matters: Willison traces four games from chat prototype to coding agent, and the process still keeps human playtesting and gameplay fixes in the loop.

The Decoder

Cathedral says its AI agent breached Anthropic via third-party software; Anthropic disputes scope

In bug-bounty research, security startup Cathedral used an AI agent to get into Anthropic's internal systems through third-party business software. Cathedral says it found a path to meeting notes, salary data and Slack messages in under twelve hours. Anthropic confirms some sales, HR and IT systems were accessed, says Claude conversations, source code and model weights were not exposed, and closed the affected access on the day it found the intrusion.

Why it matters: An AI agent reached Anthropic's internal data through a third-party integration, showing how outside services widen the reach into company systems.

AI HOT · Tips & opinions

Pydantic AI benchmark: code mode cuts long-task cost 58% and time 51%

Pydantic AI tested 8 models across 5,276 scored runs in an internal benchmark. Code mode cut long-task cost by 58%, time by 51%, and raised success rate by 16.0 percentage points.

Why it matters: The same tool set was run two ways, and code mode saves on long tasks while adding overhead on short ones.

Latent Space

How Anthropic ships Claude with Claude: splitting an agent's brain from its hands

Anthropic's engineering and product leads explain how they build their own products on shared Claude Platform components, and how they coordinate model releases, safety and scaling. Managed Agents separate session logs and the runtime framework from the code execution sandbox, so a sandbox failure doesn't kill the whole session. Service credentials are stored separately and attached through a proxy. The Opus 5.5 release also involved API integration changes to prevent knowledge distillation, and customers running evals in their own frameworks fed back anomalous behavior.

Why it matters: Anthropic separates agent sessions from execution sandboxes, treating state persistence, failure recovery and credential isolation as distinct architecture problems.