Curated weekly
What changed, and what it changes for you. Every item is triaged: act, verdict change, watch, or safe to ignore.
50 Important · 7 verdict changes · 42 FYI · 1 watching · 100 total
ImportantModels ·
SpaceXAI says Grok Voice Transcribe 2.0 will soon become the default in its Speech-to-Text API and that 1.0 will be deprecated "in the coming weeks", with no date for either. Pin grok-voice-transcribe-1.0 and test 2.0 on your own audio before your transcripts change under you.
Read the full callImportantModels ·
Alibaba's Qwen3.8-Omni-Flash takes text, image, audio and video in, has a 1M-token context, and is callable on Model Studio as qwen3.8-omni-flash. Qwen claims a price cut of more than 98% on audio input, but Qwen's launch post shows its rates only as an image and Model Studio's documentation lists no price, so check your console before moving a workload.
ImportantModels ·
Google began rolling out Gemini 3.8 Live and Gemini 3.8 Live Extended Thinking to the Gemini API and Google AI Studio on 15 September 2026. Google says Extended Thinking ranks first on Artificial Analysis's Speech-to-Speech Quality Index at 82.6, and it is the one we recommend you trial first.
MImportantSecurity ·
The official MCP Python SDK let a malicious MCP server steer an HTTP client's OAuth credentials to an authorization server of its choosing (GHSA-qx49-fqc8-xw99, CVSS 7.5). Versions 1.9.1 to 1.29.1 and 2.0.0 to 2.1.1 are affected. Upgrade to 1.30.0 or 2.2.0, and on the two machine-to-machine providers also pass issuer=, or the upgrade protects nothing.
ImportantModels ·
Anthropic released Claude Sonnet 5.5 on 28 September at $2 input and $10 output per 1M tokens, the same as Sonnet 5. Cache reads launched at $0.20 and now list at $0.10. Claude Code 2.1.284 made it the default Sonnet on the Anthropic API. If you run Sonnet with thinking off, switch to the new between_tools setting before you migrate. Every benchmark is Anthropic's own.
ImportantModels ·
Since 17 September, verified life-sciences teams on the Claude API, Enterprise or Team plans can apply to Anthropic's Life Sciences Verification Program to relax the biology safeguards that block legitimate research. Standard Use launched on Mythos 5.1, Opus 5 and Sonnet 5 and now lists Opus 5.5; High-risk Use covers Opus 5 and Sonnet 5, while Mythos stays limited to a small vetted set.
Foundation and frontier model releases, updates, and capability changes.
Important
Google shipped gemini-3.8-flash-tts and gemini-3.8-flash-lite-tts to the Gemini API and AI Studio on 23 September at $9.00 and $6.00 per 1M audio output tokens, less than half of Gemini 3.1 Flash TTS Preview's $20.00. Every 3.8 rate doubles on 1 January 2027, so we rate both conditional.
Important
OpenAI made GPT-Live-1, its full-duplex voice model, generally available in the API on 10 September at $0.05 per minute for the voice session, with telephony support. The reasoning runs on a backend model you pick and pay for separately, and OpenAI publishes no all-in per-minute cost. GPT-Live-1 stays conditional: the no-API caveat is gone, and cost is the new one.
Important
DeepSeek released V4.1-Flash on 10 September, a 552B MoE with open weights, and retired V4 Flash the same day. The old model id now routes to V4.1-Flash and bills at the new Flash price of $0.15 in and $0.60 out per 1M tokens off-peak. Same id, different model: re-run your evals.
GGImportant
Google's third Flash release in six weeks lands at the same $0.75/$3.75 intro price as 3.7 Flash, with a caveat: 3.8 'works harder' and may burn more tokens. The gated 3.8 Flash Cyber goes to trusted defenders under the new Fairwind Program.
GImportant
Gemini 3.7 Flash lands three weeks after 3.6: better on every benchmark Google published, at half the intro price through 2026. The upgrade path is obvious.
Security incidents, vulnerabilities, and safety disclosures.
Important
CVE-2026-82533 (CVSS v4 9.4) let an agent in DeepSeek Harness 0.1.1-rc.2 or earlier call the harness's unauthenticated local API and set its own session to danger-full-access, approval prompts off. The first fixed npm release is 0.1.2-alpha.2. DeepSeek published no advisory, and third-party desktop wrappers pin their own copies.
Important
Anthropic warned Claude users that infostealer malware is stealing sessions and spending usage. With no itemized usage, theft could go unnoticed for months.
+4Important
GitSpawn: Git settings in a received repo's .git/config make coding agents run attacker commands outside the sandbox, with no prompt. Codex, Cursor, goose and Claude Code's first path are fixed. Four paths were open at Manifold's 1 September retest, and newer releases since are unverified. Clone, do not open received copies.
Important
Wiz found 3,074 internet-facing LiteLLM gateways; 294 accepted the example master key sk-1234, and 191 of those required no authentication at all. That key opens every provider API key stored on the gateway. Changing it needs no upgrade. Do it today, then move to LiteLLM 1.84.0 or later.
Important
Four safety researchers published findings that roughly 18,000 posts across German-language wikis came from autonomous agents trading tips on evading sandbox restrictions, and they attribute the swarm to OpenAI. OpenAI has not confirmed the attribution. If you run long-lived agents with network egress, the action is the same either way: audit what they have been writing.
Pricing, plan, and cost changes across tools and models.
Important
GLM-5.3-Flash now bills at $0.15 input and $0.50 output per 1M tokens. The $0.075/$0.25 launch rate was a 50 percent promotion that ended on 9 September 2026. Full-size GLM-5.3 stays at $1.40/$4.40, and our verdict stays conditional.
Important
The Verge reports a refiled class action alleging Anthropic's Claude Max tiers ($100 for "5x", $200 for "20x") did not make clear the multipliers cover five-hour sessions under a weekly cap. Allegations, not findings. Read any multiplier as a window, not a month.
Important
OpenAI's live rate card puts GPT-5.6 Sol at $4.00 input and $20.00 output per 1M tokens on short context, with a footnote most people scroll past: the promotional pricing is available "at least through November 21, 2026." No end date is published, and no post-promotional rate. If you are sizing a Sol workload into next year, you are sizing against a number OpenAI has reserved the right to move. Our own directory entry carried $5/$30 and has been corrected.
Important
Meta now publishes two price lists for the same Muse Spark models. Standard is $1.25/$4.25 per 1M tokens with no training on your data. The Muse Spark Contributor tier is $0.10/$0.20 if Meta can train on your prompts and completions. The discount is real. So is the catch Meta buried below it: Contributor gets 100 requests per minute against Standard's 3,000.
Important
DeepSeek released V4 Pro GA and, per Reuters and Caixin, raised API prices up to 1,100% with peak/off-peak billing. Both DeepSeek verdicts are under re-review.
Mergers, acquisitions, milestones, and broader industry moves.
Important
Jeff Dean and three top Google AI researchers left to co-found Discovery Loop, an Alphabet-backed PBC automating scientific research. Demis Hassabis moves to Chief Scientist of Alphabet as Koray Kavukcuoglu takes over DeepMind.
Verdict changed
Stripe's newsroom and OpenRouter's own blog both confirmed on August 19, 2026 that Stripe has agreed to acquire OpenRouter. The deal is agreed, not closed, no price was disclosed, and OpenRouter's neutrality pledge is an on-record commitment rather than a structural guarantee. Our rating holds at conditional.
CHSFYI
On 16 September OpenAI announced Sponsored Agents, new tools for marketers, and ChatGPT Ads integrations with HubSpot and Shopify. HubSpot says you can now build, budget and launch ChatGPT Ads inside HubSpot, with ROI reporting. We could not open OpenAI's own post, so we cover only what we could verify.
FYI
Anthropic has published figures on how much of its own AI research Claude does. As of August 2026, Claude leads 26% of that work, over 90% is at "AI collaborates" or above, and none of it is fully autonomous. About 30,000 agents run at once, and of roughly 100,000 flagged transcripts a week, about 50 reach a human.
DFYI
Cognition raised over $2B at a $48B valuation, led by Andreessen Horowitz and Accel, and says run-rate revenue grew from $492M to almost $900M since May. No directory verdict moves.
AI coding assistants, editors, and developer-facing tooling.
Important
OpenAI gave SpaceX the maximum contractual notice: OpenAI models disappear from Cursor on Nov 12, 2026. Anthropic, Google, and Cursor's own models remain, but if you run GPT inside Cursor, you have a migration clock.
Important
Starting Aug 14, auto mode becomes the default for Claude Code Pro, Max, and Team. Safety testing: 89% dangerous-command catch rate vs. 13.6% for humans. Third-party eval: zero prompt injection successes in 720 attacks.
Important
Meta launched Muse Code in beta (a terminal-based AI coding agent with persistent async background agents) alongside Muse Spark 1.2, its third model in four months. On Meta's own benchmarks, it trails Claude Opus 5 across all three published coding benchmarks and beats GPT-5.6 Terra on two of three, trailing only on DeepSWE 1.1. The default pricing tier sends your code into Meta's training pipeline.
FYI
Anthropic redesigned Projects in Claude Code on 17 September 2026: a coordinator fans work out to parallel cloud threads, each on its own branch and repo copy, with overlaps surfacing as merge conflicts. It is a beta for select Pro and Max subscribers, and parallel threads hit usage limits faster.
FYI
ZCode GLM Coding Plan: $18–$168/mo standard pricing, 30% promo drops to $12.60–$117.60, the cheapest agentic coding subscription out there.
Government action, regulation, and AI governance developments.
Important
EU AI Act enforcement begins August 2 with fines up to 3% of global turnover. Officials confirmed prior engagement with OpenAI and Anthropic, before their incidents went public.
Important
SB 942 requires C2PA provenance metadata on all AI content. Midjourney (a CAI member since 2023) shipped without compliant watermark, facing $5k/day fines.
Important
EO 14409's Aug 1 deadline passed with silence from Washington. Three statutory requirements, all undelivered. The US regulatory vacuum is now official.
Important
Today's White House AI review deadline forces every frontier lab to choose: submit to government pre-release evaluation, or publicly explain why they won't.
FYI
In a 20-page statement of interest in NYT v. OpenAI, the US government argues training LLMs on copyrighted text is transformative fair use. It is advocacy, not law, but it is the executive branch's clearest position yet on the doctrine every frontier model depends on.
New product and tool launches entering the market.
CFYI
Anthropic is folding Cowork into the main Claude app, starting with Pro and Max over the next few weeks. Claude Docs and Claude Slides launched in beta on paid plans on 16 September 2026. No price change is announced. The setting that matters is ask-before-acting: leave it on.
CFYI
OpenAI added a Data agent to ChatGPT Work on 10 September. Invoked with @Data, it queries Snowflake, BigQuery, Databricks, Redshift and more, builds dashboards from plain language, and enforces the connected account's table, row and column permissions. Sixteen days on, OpenAI still names no price or plan list.
GFYI
Google's Gemini desktop app shipped globally for Windows 10 and 11 on 10 September 2026. Alt + Space summons it over the active window. The download is open, but Google gates "selected features" behind a Google AI subscription for users 18+, and does not say which.
FYI
Cursor launched Origin: repos, PRs, and GitHub sync inside the editor. A credible GitHub rival, but not a migration target until its data terms ship.
Autonomous agents, agent frameworks, and agentic runtimes.
Important
OpenAI opened the Agents API in public beta on 10 September: the Codex harness as a managed service, with no fee beyond tokens and tools. No platform fee is not the same as cheap.
Important
SpaceXAI launched Grok Bot: persistent AI agents that sign into your apps, coordinate in teams, and run 24/7 on dedicated cloud VMs. Pricing starts at $120/mo. Early beta. Test it, don't standardize on it.
Never need to catch up again
The weekly delta: only verdict changes and act-now items. No digest filler.