Nothing fails until you switch models. Before switching, search your code for "disabled" thinking and forced toolchoice, and test any flow that moves a conversation between accounts.

29 Sept 2026Environments & integrations

GitHub Enterprise Cloud enforces the self-hosted runner minimum today, 29 September

Teams that run coding agents or CI on self-hosted runners with pinned images. Check runner versions this morning; the new runner-deprecation REST API returns the cut-off dates for each version.

29 Sept 2026Security & privacy

MCP TypeScript SDK 2.2.0 and 1.31.0 bind OAuth credentials to their issuer

Anyone running MCP clients against servers with long tool lists, or keeping OAuth tokens in a strict schema.

28 Sept 2026Environments & integrations

LiteLLM 1.103.0 deprecates the `litellm-proxy` command. Scripts that call it should switch to `lite`.

Nothing fails today. Replace litellm-proxy with lite in scripts, Dockerfiles and CI now, and check any job that treats stderr output as a failure.

28 Sept 2026Environments & integrations

worktrunk 0.80.0 runs commit hooks on a detached HEAD during a squash

Anyone running worktrunk to manage agent worktrees with commit hooks that read the current branch, or scripts that parse its hook logs.

25 Sept 2026Environments & integrations

Claude Code 2.1.282 ignores telemetry-export variables in project settings

Teams that collect Claude Code usage through OpenTelemetry configured per repository.

25 Sept 2026Security & privacy

GitHub Copilot Business and Enterprise admins have until 22 October to choose what new Copilot features do by default.

Pick a default this week, and set MCP servers and Code Review explicitly so neither depends on the global setting.

24 Sept 2026Agent workflows

Claude Code skips AGENTS.md whenever it cannot fetch a remote flag, and nothing on Anthropic's pages says that is fixed.

Do not rely on native AGENTS.md loading yet. Put a one-line CLAUDE.md containing @AGENTS.md next to it; @ imports do not depend on the flag.

24 Sept 2026Environments & integrations

Node 20 is gone from GitHub Actions runners, and so is the opt-out

Anyone pinning an old third-party action by SHA, or running self-hosted runners on older Macs or ARM32 boards.

24 Sept 2026Security & privacy

The GitHub Copilot app can sandbox local sessions per project

Teams letting an agent run shell commands against a working tree that also holds credentials.

23 Sept 2026Cost & efficiency

Anthropic and OpenAI both cut prices on the same Tuesday, and Anthropic's cut arrives attached to three migration errors.

Worth moving to, but read the migration notes before you change the model string. The three 400s are the whole of the work, and each has a named replacement.

23 Sept 2026Security & privacy

GitHub is removing SHA-1 RSA SSH signatures and a Diffie-Hellman key exchange, on a dated schedule

Anyone whose CI, deploy keys or agent sandboxes speak Git over SSH from a pinned or old client image.

23 Sept 2026Cost & efficiency

GPT-6 prompt caching gets a 30-minute reuse window, effort changes that survive the cache, and a miss diagnostic

Anyone running a long-lived agent loop on GPT-6 whose input bill is mostly re-sent context.

23 Sept 2026Cost & efficiency

GPT-6 Sol and Luna ship at half their predecessors' prices

Anyone with a Sol or Luna line in a production bill, and anyone who priced an agent out at the old rate.

22 Sept 2026Security & privacy

Cline CLI 3.0.63 closes a Windows binary-hijack path and stops subagents re-asking for approval

Anyone running the Cline CLI on Windows, and anyone who has delegated work to configured subagents.

22 Sept 2026Cost & efficiency

Grok 4.7 is rolling out in GitHub Copilot, billed at provider list pricing

Copilot Business and Enterprise admins, whose default-enablement policy turns new models on unless they opted out.

22 Sept 2026Environments & integrations

vLLM 0.30.0 deletes four things you may be running and shortens two model context windows without asking.

Read the upgrade notes before this reaches a pinned deployment. The removals are recoverable; the YaRN context-length change can surface as user-visible request failures instead of a startup error.

21 Sept 2026Tools & models

The loudest thing on the internet today is a v0.3.0 alpha that needs a second alpha underneath it.

go install the CLI and read the concepts doc if the density argument interests you. Nothing here survives a migration plan yet.

21 Sept 2026Environments & integrations

Worktrunk 0.79.0 changes what `--no-cd -x` runs against

Anyone who has wt switch --no-cd -x inside an alias, a hook or a script.

18 Sept 2026Environments & integrations

`ubuntu-latest` moves to 26.04 between 19 October and 19 November

Every repository whose CI says ubuntu-latest — which is most of them.

18 Sept 2026Agent workflows

Cline 4.1.19 stops guessing when to compact

Anyone pointing Cline at minified bundles, binaries or screenshots and watching long tasks die quietly.

18 Sept 2026Environments & integrations

Cline desktop 0.0.31 drives a remote machine over SSH

Anyone who wants an agent's file writes and shell to happen on a dev box rather than a laptop.

18 Sept 2026Environments & integrations

Two harnesses shipped a build that would not work, and fixed it the same night.

On Claude Code 2.1.275 behind a gateway, or Cline desktop 0.0.31: you are broken, and the fix is published. Upgrade both.

17 Sept 2026Cost & efficiency

HarnessTax priced 21 model-harness pairs. The harness moved the bill further than the score.

The design and the raw chart data are public and worth an hour. They will not settle a harness migration on thirty tasks — price your own rollouts before you move.

17 Sept 2026Security & privacy

pydantic-ai 2.44.0 and 1.107.6 patch four advisories

Anyone running pydantic-ai agents with webfetch enabled, or shipping OTel spans off-box.

17 Sept 2026Environments & integrations

worktrunk v0.78.0 renames every hook context key

worktrunk users with hook scripts or --execute templates checked into a shared repo.

16 Sept 2026Tools & models

Gemini 3.8 Live and 3.8 Live Extended Thinking, in the API today

If you have been stitching a cascaded stack out of a recogniser, a model and a voice, this collapses three hops into one, with Live API integrations through LiveKit, Pipecat and Vercel. Enterprise access is still private preview.

16 Sept 2026Security & privacy

GitHub has disabled SHA-1 in HTTPS

Anyone whose pipeline still builds on an old base image. A TLS stack that only offers SHA-1 signature algorithms can no longer reach github.com, and it surfaces as a handshake failure in a clone that worked on Monday. Audit the oldest image in your CI first.

16 Sept 2026Cost & efficiency

TypeSafe prices Jev at $0.042 per million input tokens with free output, then gates it behind a waitlist.

Join the waitlist and keep the pricing table. Hold the verdict until somebody outside the company has run it on their own workload.

15 Sept 2026Security & privacy

Claude Code v2.1.271 · per-command allowed_domains, plugin commands pinned by hash

Anyone running Claude Code under an organisation policy, or installing plugins from a script

15 Sept 2026Security & privacy

Cline's hub loads agent plugins from your home directory and deliberately refuses to load them from the repository you just opened.

Upgrade. On a large workspace, or on a provider that forwards 429s, this removes two failure modes you may have been blaming on your own machine.

15 Sept 2026Tools & models

llama.cpp v0.4.1 · a signature changes, and GDN normalisation is fixed

Anyone compiling against libllama, and anyone serving Qwen, Kimi or GLM locally

14 Sept 2026Environments & integrations

Cline desktop 0.0.27 · sign-in errors, sign-out, OpenCode Go routing

Anyone already running the Cline desktop app, especially on OpenCode Go models or Cline Pass

14 Sept 2026Environments & integrations

GitHub ran a critical incident for 1h 29m on Sunday morning. The rest of the window produced one release.

Nothing to install beyond Cline desktop 0.0.27. Re-run any GitHub-dependent job that failed between 09:16 and 10:45 UTC on Sunday before you spend time debugging it.

11 Sept 2026Environments & integrations

ADK Python 2.9.0 · failover + LiveKit; resume and GCS paths break

Teams on google/adk-python with workflows, GCS tools, or MCP servers

11 Sept 2026Environments & integrations

Claude Code 2.1.268 · third-party endpoint 400 fix + gateway pricing

Anyone on Claude Code latest/next — mandatory if ANTHROPICBASEURL or a third-party compatible endpoint is in the path

11 Sept 2026Environments & integrations

Claude Code 2.1.268 fixes third-party endpoint 400s since 2.1.265; ADK Python 2.9.0 breaks workflow resume and GCS paths; MAI-Code-1-Flash is off Copilot.

Claude Code: pin ≥2.1.268, especially on ANTHROPICBASEURL / third-party gateways. ADK operators: read 2.9.0 before upgrade — make node bodies idempotent and set localfileroot. Expect MAI-Code-1-Flash gone from Copilot.

10 Sept 2026Environments & integrations

Claude Code 2.1.267 · effort cap + cache/resume fixes

Anyone on Claude Code latest/next — especially gateway, Bedrock/Vertex, or long resume sessions

10 Sept 2026Tools & models

Codex 0.154 puts GPT-6 Astra in the picker; Claude Code pins to 2.1.267; MAI-Code-1-Flash leaves Copilot today.

Codex users who want Astra: ≥0.154.0. Claude Code: pin ≥2.1.267. Expect MAI-Code-1-Flash gone from Copilot. Copilot Business/Enterprise admins: review agent permission policies.

10 Sept 2026Environments & integrations

Codex 0.154.0 · Astra in picker + experimental worktrees

Anyone on Codex CLI/TUI who wants Astra or isolated session checkouts

10 Sept 2026Security & privacy

Copilot · enterprise managed agent permissions (GA)

GitHub Copilot Business/Enterprise admins rolling out agent workflows

9 Sept 2026Environments & integrations

Claude Code 2.1.266 · gateway regression fix

Anyone on Claude Code with a gateway, proxy, or non-default auth path — and anyone who landed on 2.1.265 overnight

9 Sept 2026Environments & integrations

Claude Code 2.1.266 undoes a 2.1.265 gateway sign-in break for proxy setups.

Pin Claude Code ≥2.1.266. If you took 2.1.265 on a gateway/proxy path, upgrade today. MAI-Code-1-Flash leaves Copilot tomorrow (10 Sep).

8 Sept 2026Security & privacy

MCP Python SDK 2.2.0 · origin redirects, 30 min idle, OAuth issuer

Anyone hosting or calling MCP over HTTP with the official Python SDK

8 Sept 2026Security & privacy

MCP Python SDK 2.2.0 tightens redirects, idles sessions at 30 min, and checks OAuth issuer.

Read 2.2.0 before upgrading MCP Python 2.x — redirects, idle sessions, and OAuth issuer checks change defaults. 1.x operators: ≥1.30.0. MAI-Code-1-Flash leaves Copilot 10 Sep.

7 Sept 2026Tools & models

Quiet window: zero watchlist releases to 06:00 UTC 7 Sep; hold Friday’s pins.

No new decision pin inside the window. If you track npm latest, 2.1.263 is the tag; stable remains 2.1.236. MAI-Code-1-Flash leaves Copilot 10 Sep.

4 Sept 2026Cost & efficiency

Claude Code 2.1.260 · Fable cache + effort

Anyone on Fable 5.1 in Claude Code after 2.1.257–259

4 Sept 2026Cost & efficiency

Claude Code 2.1.260 fixes Fable 5.1 cache and effort mid-session; npm latest is 2.1.260.

Pin Claude Code ≥2.1.260 if you run Fable 5.1. On Copilot, enable Gemini 3.8 Flash and plan the 2 Oct model cutover.

4 Sept 2026Environments & integrations

Cline Desktop 0.0.23 · Agent Plugins via Hub

Cline Desktop users and anyone with both desktop and CLI installed

4 Sept 2026Tools & models

Copilot · Gemini 3.8 Flash in; four models out 2 Oct

Copilot Pro through Enterprise pinning Gemini Flash, Kimi, or Opus 4.7

3 Sept 2026Security & privacy

Claude Code 2.1.259 · managed MCP + headless

Org Claude Code fleets and unattended hosts

3 Sept 2026Environments & integrations

Codex 0.153.0 · plugins, Vim undo, reconnect

Codex CLI/desktop on 0.152.x

3 Sept 2026Environments & integrations

Codex 0.153.0 ships; Claude Code latest is 2.1.259 with managed MCP and headless deny.

Pin Codex ≥0.153.0 over 0.152.x. Claude Code latest 2.1.259 for multi-session and headless; stable only if you need the slow channel.

3 Sept 2026Environments & integrations

Copilot · content exclusions GA; enterprise default model

Copilot Business/Enterprise admins shipping agentic app/CLI

2 Sept 2026Environments & integrations

Claude Code 2.1.258 · Monterey + remote session fixes

Mac 12 hosts and anyone on remote/scheduled Claude Code

2 Sept 2026Tools & models

Claude Fable 5.1 · default Fable, Copilot GA

Long-running agent work on Claude Code, API, or Copilot

2 Sept 2026Tools & models

Claude Fable 5.1 is the default Fable model — in Claude Code, on the API, and in Copilot.

Pin Claude Code ≥2.1.258. Try Fable 5.1 for long agent runs; on Copilot, read the retention policy before enabling org-wide.

2 Sept 2026Tools & models

Cline 4.1.17 · Hub memory balloon + Fable default

Cline extension/desktop on long tasks or unpinned Anthropic models

1 Sept 2026Environments & integrations

Claude Code 2.1.252 · four reliability fixes

Mac CLI and Remote Control hosts on latest

1 Sept 2026Environments & integrations

Codex 0.152.0 · stable after alphas

Codex CLI/desktop on 0.151 or a 0.152 alpha

1 Sept 2026Environments & integrations

Codex 0.152.0 is stable; Claude Code latest is 2.1.252.

Pin Codex ≥0.152.0. Claude Code latest 2.1.252 for Mac/Remote Control fixes; stable only if you need the slow channel.

31 Aug 2026Tools & models

Quiet Monday. No watchlist release that changes a pin.

No pin from today. Plan Cursor OpenAI traffic through 12 Nov; re-budget Claude Code weekly after 14 Sept.

28 Aug 2026Environments & integrations

Anthropic Console · personal and service-account API keys; Python SDK 1.2.0 GA shapes

orgs rotating API keys and anyone still on files-api / skills beta headers

28 Aug 2026Security & privacy

Claude Code 2.1.248 · --restricted, hourly cache fix, cross-session messaging

anyone running Claude Code in CI, shared machines, or multi-hour sessions

28 Aug 2026Security & privacy

Claude Code 2.1.248 adds --restricted; npm latest is 2.1.250 while stable stays 2.1.236.

latest ≥2.1.248 for --restricted and the hourly cache fix; pin stable only if you need the slower channel.

27 Aug 2026Agent workflows

Codex 0.150.0 · @task mentions, Interrupt hooks, /copy picker

terminal Codex users and teams wiring interrupt-time automation

27 Aug 2026Environments & integrations

vLLM 0.28.0 · Runner V2, tiered KV offload, bitsandbytes out-of-tree

anyone running vLLM servers or pinning CUDA/ROCm serving images

27 Aug 2026Environments & integrations

vLLM 0.28.0 ships Model Runner V2 defaults and breaks bitsandbytes-in-tree; Claude Code latest is 2.1.247 while stable stays 2.1.231.

vLLM: checklist bnb plugin + Transformers ≥5.15 before upgrade. Claude Code: latest ≥2.1.247 for those fixes; stable remains 2.1.231.

26 Aug 2026Cost & efficiency

ChatGPT Plus gets the Codex and Work 5-hour cap back — Pro $100/$200 stay uncapped for now.

Plus Codex/Work: plan around a 5h rolling window again, or move Pro / credits / another harness. Claude Code on Bedrock or a gateway: stay on latest ≥2.1.239 (ideally 2.1.246); stable still lacks those fixes.

26 Aug 2026Environments & integrations

Claude Code 2.1.246 · MCP honesty, background sessions, gateway edge cases

anyone on latest/native Claude Code, MCP servers, or ANTHROPICBASEURL gateways

26 Aug 2026Tools & models

llama.cpp 0.3.0 · dots3-note multimodal, DeepSeek 4 tensor-split, ggml 0.22

local llama.cpp / server / multimodal runners, especially DeepSeek 4 multi-GPU

25 Aug 2026Cost & efficiency

GPT-5.6 Sol is $4/$20 per 1M tokens — −20% input, −33% output — promo through at least 21 Nov 2026.

If you already call Sol (or gpt-5.6), re-check invoices and budget alerts against $4/$20. Prefer Batch or Flex for offline work. Pin gpt-5.6-sol if you need the promo rate to stay attached to a snapshot name.

25 Aug 2026Environments & integrations

MCP Python SDK 2.1.0 · quieter handler errors, leaner content returns

anyone shipping MCP servers or clients on the Python SDK

24 Aug 2026Environments & integrations

Claude Code stable is still on 2.1.231 — nine releases behind latest, missing two billing fixes.

On Bedrock or a gateway: install ≥2.1.239 (or set minimumVersion 2.1.239 on stable). Confirm with claude --version. Others can wait for stable.

24 Aug 2026Environments & integrations

Cline 4.1.14–15 · catalog refresh + MCP auto-approve fix

Cline users with MCP tools or who need the new catalog rows

21 Aug 2026Environments & integrations

anthropic Python SDK hits 1.0 — httpx2, Python 3.10+, Text Completions gone.

Pin anthropic>=1.0.0 only after you clear MIGRATION.md. If you construct HTTP clients or mock httpx, switch imports to httpx2 first. Drop any remaining Completions callers. Python <3.10 stays on 0.125.x.

21 Aug 2026Agent workflows

Codex rust-v0.149.0 · agents dashboard, queue, cwd commands

Codex multi-session operators and SDK embedders

21 Aug 2026Environments & integrations

pydantic-ai 2.33.0 · emergency Anthropic 1.0 pin

anyone on pydantic-ai with Anthropic models and unpinned anthropic

20 Aug 2026Environments & integrations

Anthropic Files API and Agent Skills are GA — drop the beta headers.

Pin anthropic ≥0.124.0, drop both beta headers on new code, and set file expiry on upload. Keep the beta header only if a client still parses the pre-GA list shape.

20 Aug 2026Environments & integrations

Claude Code 2.1.236–237 · default model env, idle notify, Concise style, gateway cache

Claude Code daily drivers on multi-session hosts, gateway proxies, and anyone drowning in preamble

20 Aug 2026Security & privacy

openai-agents-python 0.22.0 · guardrail redaction + explicit-client contract

OpenAI Agents SDK users with guardrails, multi-checkpoint runs, or explicit client wiring

19 Aug 2026Agent workflows

Claude Cowork is on web and mobile for every paid plan — still beta.

Turn it on for Pro/Max/Team if you want cross-device handoff. Keep auto-approve off for mail and Drive until you have tried a throwaway folder. Enterprise: confirm the admin toggle before assuming the rollout reached you.

19 Aug 2026Agent workflows

Codex rust-v0.148.0 · /export, session fork, Bedrock Runtime, thread cost

Codex CLI/TUI users on multi-hour sessions, Bedrock shops, and anyone metering agent spend

19 Aug 2026Environments & integrations

openai-python 3.3.0 · named data-residency endpoints

teams that must pin OpenAI calls to a residency region from Python

18 Aug 2026Security & privacy

Claude Code 2.1.234 · auto-continue at usage-limit reset + NTLM path harden

Claude Code daily drivers on Pro/Max/Team and multi-session hosts

18 Aug 2026Agent workflows

Cursor Origin early beta is rolling out to paid plans — repos, PRs, and GitHub sync inside Cursor.

Treat Origin as a research preview: try on a throwaway org if you want agent-side repo UX, keep production source of truth on GitHub until data terms and agent-native merge features land. Enterprise admins who never decided on a second host should confirm the opt-out posture.

18 Aug 2026Environments & integrations

openai-python 3.2.0 · Bedrock Runtime endpoint

teams calling OpenAI-format APIs through Bedrock Runtime

18 Aug 2026Environments & integrations

pydantic-ai 2.31.1 · Bedrock structured-output deny for Sonnet 5 / Fable 5

pydantic-ai agents on Bedrock or Gemini with strict schemas

17 Aug 2026Environments & integrations

DeepSeek Harness · open developer preview

harness authors comparing open runtimes to Claude Code / Codex

17 Aug 2026Cost & efficiency

DeepSeek peak/off-peak API rates are live as of 16 Aug 16:00 UTC.

Meter DeepSeek by UTC hour. Prefer off-peak for batch and long loops; pin Pro-0813 or Flash-0731 so a silent version cell cannot surprise you.

17 Aug 2026Security & privacy

openai-agents-python 0.21.1 · model call timeouts + sandbox controls

OpenAI Agents SDK users who need hard call budgets or tighter sandboxes

14 Aug 2026Security & privacy

Claude Code auto mode is the default for new Pro, Max, and Team sessions as of today.

If you never set a default, decide before the next session: keep auto and pin ask rules on git push / gh pr create, or set "defaultMode": "manual" in ~/.claude/settings.json. Org admins who want the old baseline set permissions.disableAutoMode to disable.

14 Aug 2026Cost & efficiency

DeepSeek API · peak/off-peak prices from 16 Aug 16:00 UTC

anyone metering DeepSeek as a default agent model

14 Aug 2026Tools & models

GitHub Copilot · Gemini 3.7 Flash rolling out

Copilot orgs picking a cheap coding model this week

14 Aug 2026Environments & integrations

google-adk 2.7.0 · model-declared capabilities + media tools

ADK agents that pair output schemas with tools, or return media from tools

14 Aug 2026Security & privacy

pydantic-ai 2.30.0 / 1.107.5 · Host-header fix on local web UI

anyone who runs the pydantic-ai dev web chat on a workstation

13 Aug 2026Environments & integrations

Agent Plugins 1.0 · GA in VS Code, Copilot CLI, Copilot app

plugin authors and Copilot org admins

13 Aug 2026Security & privacy

Claude Code 2.1.229 · auto mode fix + marketplace command sources

auto-mode users without attribution headers; plugin authors

13 Aug 2026Environments & integrations

Cline SDK 0.0.74 + Desktop 0.0.12 · Claude Code provider fixed

Cline users on Claude Code or desktop

13 Aug 2026Tools & models

DeepSeek V4 Pro leaves preview as 0813; Grok 4.6 ships the same day for long-running agents in Cursor and Grok Build.

Pin Pro 0813 and run one trusted agent loop; budget for a price change with no date. Try Grok 4.6 on one long job this week under the 2× promo; keep Composer for cheap daily edits.

13 Aug 2026Cost & efficiency

goose 1.46.0 · unrolled loop, cache-safe turns, per-message cost

goose hosts metering multi-turn runs

12 Aug 2026Tools & models

Copilot · MAI-Code-1.1-Flash in, 1-Flash out 10 Sept

Copilot teams on MAI-Code-1-Flash or picking a cheap coding model this month

12 Aug 2026Agent workflows

Copilot for JetBrains · memory + Ollama BYOK

JetBrains users who want local models or cross-session memory without leaving the IDE

12 Aug 2026Tools & models

Ollama 0.32.9 · Nemotron 3.5 Lightning

local-agent hosts already on Ollama who want a small-active MoE next to Glimmer

12 Aug 2026Environments & integrations

openai-python 3.0.0 · HTTPX2 default

anyone who constructs a custom httpx client for the OpenAI SDK

12 Aug 2026Security & privacy

pydantic-ai 2.28.0 · web UI CSRF fix

anyone who runs the pydantic-ai dev web chat on a workstation

12 Aug 2026Agent workflows

xAI ships Grok Bot beta: always-on agents with their own cloud computer, on SuperGrok Heavy and Cursor Ultra/Teams Premium.

real product on paid tiers. Try one end-to-end job on your own tools before you trust the seat. Skip if you are not already on Ultra or Heavy.

11 Aug 2026Cost & efficiency

Claude Sonnet 5 · $2/$10 made permanent

anyone whose agents default to Sonnet 5 and who had a 50% step-up in the September forecast

11 Aug 2026Environments & integrations

Cline v4.1.8

Cline users who thought Yolo did something, and anyone on Vertex

11 Aug 2026Tools & models

Meta ships Muse Glimmer: 30B, Apache 2.0, built to run local agents on one GPU.

weights and HF load path are real; treat consumer-GPU speed claims as vendor until you time a tool loop on your box. Pin a digest, run one agent scaffold you already trust, measure tokens/sec and tool-call success before you move anything off the cloud default.

11 Aug 2026Environments & integrations

OpenAI Agents Python 0.20.0

anyone who depended on the previous default model without setting one

11 Aug 2026Security & privacy

Vercel Sandbox · Managed Images

anyone spinning agent sandboxes on Vercel who was baking tools at boot

11 Aug 2026Environments & integrations

vLLM 0.27.0

self-hosters pinning a serving engine this week, especially Kimi K3 or DeepSeek-V4 traffic

10 Aug 2026Environments & integrations

AI SDK · openai-compatible 3.0.28

anyone metering usage behind an OpenAI-compatible proxy serving reasoning models

10 Aug 2026Security & privacy

Claude Code makes auto mode the default on 14 Aug. The number that sold it: humans caught 13.6% of planted dangerous commands.

already on auto mode — only the free classifier is new. On manual or YOLO, decide before the 14th: Shift+Tab still flips modes; admins can pin or disable auto. Do not read 89% as solved; eleven percent still gets through.

7 Aug 2026Environments & integrations