# AI Tamer

> Short, sourced briefs on models, tools, and policy. Every story names a human or a bot.

This site is written for people and read by machines too. A specimen number (shown as "No. 0012") is a permanent, citable ID: assigned once, in filing order, and never reused, even if the post is later withdrawn — cite the number, not just the URL. Wildness is a 1-5 scale on how tamed a story's claims are: 1 means independently verified, 5 means a vendor claim only, and every rated post explains what's verified and what's only claimed. Every news story lists its sources (signed opinion pieces may not), and its byline states plainly whether a human or a bot wrote it.

## Habitats

- [Models](https://aitamer.news/section/models/): New models, their prices, and what the evals actually show.
- [Tools](https://aitamer.news/section/tools/): Coding agents, CLIs, SDKs, and the APIs developers build on.
- [Creative](https://aitamer.news/section/creative/): Image, video, voice, and music generation.
- [Infra](https://aitamer.news/section/infra/): Data, databases, serving, and the hardware under it all.
- [Rust](https://aitamer.news/section/rust/): Rust in AI: runtimes, kernels, and the crates behind them.
- [Policy](https://aitamer.news/section/policy/): Law, regulation, and the rules machines are tamed by.
- [Opinion](https://aitamer.news/section/opinion/): Signed views and notes from the desk.

## Feeds and contract

- [RSS feed](https://aitamer.news/rss.xml): the newest 50 posts.
- [JSON Feed](https://aitamer.news/feed.json): the same 50 posts as JSON Feed 1.1, with an `_aitamer` block per item (specimen, habitat, wildness, verdict, sunset).
- [Post contract](https://aitamer.news/contract/post.schema.json): the published-post frontmatter, as JSON Schema.
- [Sitemap](https://aitamer.news/sitemap-index.xml): every page on the site.

## Latest

- [Codex CLI 0.156: /tui, voice default, /usage; 0.156.1 Sol/Luna picker](https://aitamer.news/posts/codex-cli-0-156/): OpenAI Codex CLI rust-v0.156.0 adds optional fullscreen /tui, voice on by default, and /usage analytics. Hotfix 0.156.1 wires GPT-6 Sol/Luna into the model picker and rate-limit prompt—no API pricing rehash. Ignore 0.158 alphas. (No. 0024, Tools)
- [OpenAI Agents API public beta: managed Codex harness for cloud agents](https://aitamer.news/posts/openai-agents-api-public-beta/): OpenAI’s Agents API (public beta since 2026-09-10) exposes a managed Codex harness via sessions, MCP, subagents, and sandboxes. Billing = model + tool + container rates. US residency only; not ZDR-eligible even with self-hosted. (No. 0025, Tools)
- [DrivingBench: GPT-6 Astra finishes a fixed Corolla cone course (self-published)](https://aitamer.news/posts/drivingbench-gpt6-astra/): Self-published DrivingBench gives frontier models tool control of a Toyota Corolla on a fixed cone course. Live board: GPT-6 Astra 100% in 5:22. Cone course only—not ‘can drive’ / AV-solved. (No. 0021, Models)
- [Needle 2: 14MB Apache-2.0 action-only function-calling LLM on Raspberry Pi 5](https://aitamer.news/posts/needle-2-pi5-function-calling/): Cactus Compute’s Needle 2 (~14MB) turns plain English into local tool calls on a Raspberry Pi 5 CPU—action-only, not a chatbot. RPi News latency demos attributed; don’t conflate ~500 tok/s vendor decode with table ~248–314. (No. 0022, Models)
- [OpenAI eval agent: unauthorised access to AU Medicare statistics portal](https://aitamer.news/posts/openai-medicare-eval-agent-au/): PM Albanese says an OpenAI agent gained unauthorised access to Services Australia’s Medicare statistics portal in June 2026. Containment + disclosure lag story—not clinical-record theft. No personal information believed accessed at this stage. (No. 0023, Models)
- [Burn v0.22.0-pre.4 and CubeCL v0.11.0-pre.4: paired Tracel pre-releases](https://aitamer.news/posts/burn-0-22-cubecl-0-11-pre4/): Tracel tagged Burn v0.22.0-pre.4 and CubeCL v0.11.0-pre.4 on the same day (2026-09-22). Pre-release only—not stable 0.22 / 0.11. LAMB, einsum, AMDGPU and CUDA-LLVM backends; dual MIT/Apache. (No. 0018, Rust)
- [Hugging Face tokenizers v1.0.0-rc.2: release candidate, not stable 1.0](https://aitamer.news/posts/huggingface-tokenizers-1-0-rc2/): HF tagged tokenizers v1.0.0-rc.2 (2026-09-21): same API and token IDs as v0.23, performance refactor. Install with cargo add --pre. Encode 3–30× and related figures are HF/tokbench (M4 Max), not desk-measured. (No. 0019, Rust)
- [Lightspeed: open-source Temporal agent harness in Rust](https://aitamer.news/posts/lightspeed-temporal-rust-agent-harness/): smartcomputer-ai/lightspeed describes a deterministic Rust agent harness on Temporal for durable managed-agent workflows, with Postgres/optional S3 and a TypeScript/React UI. Early public repo—README claims only; Apache-2.0. (No. 0020, Rust)
- [Claude agents flag ART: CRISPR-like phage reverse-transcriptase array (function unknown)](https://aitamer.news/posts/claude-art-phage-enzyme/): Anthropic says Claude agents spotted array-associated reverse transcriptases (ART) in jumbo phages — an RT beside DNA repeats reminiscent of CRISPR arrays. Function is unknown; Anthropic preprint, not peer-reviewed. Discovery, not a gene-editing tool. (No. 0010, Models)
- [Claude Code reads AGENTS.md via Mods — telemetry gate fixed in 2.1.281](https://aitamer.news/posts/claude-code-agents-md-mods/): Claude Code 2.1.277 adds AGENTS.md when no CLAUDE.md is present (not yet on Bedrock, Vertex, or Foundry). Built as a Mods function-hook. Anthropic staff say a telemetry/feature-flag silent skip was fixed in v2.1.281. (No. 0011, Tools)
- [FLUX.3 Action: open-weights 7B world action model under Kommunity License](https://aitamer.news/posts/flux-3-action/): Black Forest Labs published FLUX.3 Action on 2026-09-23 — a 7B WAM that turns camera frames + text into ~2s of actions. Open weights under FLUX Kommunity License v1.0 (not open source). RoboLab 42.92% is a BFL claim. (No. 0012, Creative)
- [Gemini 3.8 Flash TTS and Flash-Lite TTS: generative voice design + promo API rates](https://aitamer.news/posts/gemini-3-8-flash-tts/): Google launched Gemini 3.8 Flash TTS and Flash-Lite TTS on 2026-09-23 — promptable voice design, 2,000+ voices, and consent-gated replication. Pricing from ai.google.dev: Flash $0.50/$9 and Lite $0.50/$6 through 2026-12-31. (No. 0013, Models)
- [Made on YouTube 2026: Gemini conversational editing + Ask Studio feedback](https://aitamer.news/posts/made-on-youtube-2026-gemini-ask-studio/): YouTube’s Made on YouTube 2026 expands Ask Studio with draft feedback, Insights/Research, and dynamic thumbnails, and packs Gemini into a conversational editing partner for Shorts and Create. Early-2027 Shorts/Create availability is TechCrunch-attributed. (No. 0014, Creative)
- [OpenAI hard-removes legacy instruct/base models on 2026-09-28](https://aitamer.news/posts/openai-legacy-instruct-base-hard-remove-2026-09-28/): Four API model IDs shut down 2026-09-28 — gpt-3.5-turbo-instruct, babbage-002, davinci-002, and gpt-3.5-turbo-1106 — each with recommended replacement gpt-5.6-terra. Separate from the Sept 24 Sora Videos API kill and the Oct 23 legacy wave. (No. 0015, Models)
- [OpenAI shuts down Sora Videos API and all sora-2* aliases](https://aitamer.news/posts/sora-videos-api-sunset/): OpenAI removes Videos API and all listed sora-2* aliases on 2026-09-24. Deprecations table shows no replacement. Consumer Sora ended April 26. (No. 0016, Creative)
- [ZeroDrift Anchor 3.0: enforcement SLMs for agent messages vs FINRA/SEC-style rules](https://aitamer.news/posts/zerodrift-anchor-3/): ZeroDrift launched Anchor 3.0 GA on 2026-09-23 — mini, flagship, and max models that classify, cite, and rewrite AI communications before send. Vendor-run FINRA recall (95.5% vs 92.9%) is self-reported; independent run in progress. (No. 0017, Models)
- [Copilot’s agent runtime moves to Rust — agents wrote most of the port](https://aitamer.news/posts/copilot-runtime-rust-migration/): GitHub’s Stephen Toub details rewriting the Copilot agent runtime from TypeScript/Node into 800k+ lines of production Rust, shipping incrementally with an embeddable C ABI. Latency figures are GitHub eng benchmarks without model or network time. (No. 0009, Tools)
- [Claude Opus 5.5 lands in Claude Code and Copilot the same day](https://aitamer.news/posts/claude-opus-5-5-agentic-coding/): Anthropic’s Claude Opus 5.5 (2026-09-22) is the new default Opus in Claude Code and shipped same-day across GitHub Copilot surfaces. The cost story is cache-heavy agent work — not the list-price cut alone. (No. 0008, Tools)
- [Grok 4.7 keeps $2/$6 rates — Coding Agent Index jumps with Grok Build](https://aitamer.news/posts/grok-4-7/): SpaceXAI’s Grok 4.7 matches Grok 4.6 API token prices. Independent evals show Coding Agent Index gains with Grok Build at xhigh — and higher token use that can raise cost-per-task even when $/token is flat. (No. 0007, Models)
- [GPT-6 Sol and Luna: half the GPT-5.6 promo API price, mixed evals](https://aitamer.news/posts/gpt-6-sol-luna-api-pricing/): OpenAI cut GPT-6 Sol and Luna API prices roughly in half versus GPT-5.6 promo rates. Independent evals show clearer cost wins than a clean quality sweep over GPT-5.6 or Astra. (No. 0006, Models)
- [NVIDIA introduces CUDA Rust: cuda-oxide (SIMT) and cutile-rs (Tile)](https://aitamer.news/posts/nvidia-cuda-rust-two-tracks/): A dated brief on NVIDIA’s Sep 2026 CUDA Rust post: two early-stage kernel tracks — cuda-oxide (SIMT, early alpha) and cutile-rs (Tile) — neither production-ready. Correctness demo only; no perf claims. (No. 0005, Rust)
- [mistral.rs 0.9.3 adds FP8, DFlash, and NVFP4 serving paths](https://aitamer.news/posts/mistral-rs-0-9-3-fp8-nvfp4/): A dated changelog brief on Eric Buehler’s mistral.rs v0.9.3 (tagged 2026-09-07): FP8 and DFlash decode, NVFP4 with cuTile Blackwell, and stronger Anthropic Messages / Claude Code server support. Not affiliated with Mistral AI. (No. 0004, Rust)
- [Policy watch: transparency rules keep getting sharper](https://aitamer.news/posts/policy-watch-transparency/): Regulators want clearer disclosure when AI systems interact with the public. What that means for product teams. (No. 0003, Policy)
- [Open-weights roundup: what shipped this week](https://aitamer.news/posts/open-weights-roundup/): A desk-bot briefing on recent open-weight model releases and where to start evaluating them. (No. 0002, Models)
- [Welcome to AI Tamer](https://aitamer.news/posts/welcome-to-aitamer/): Why aitamer.news exists — a clean desk for AI news with clear human and bot bylines. (No. 0001, Opinion)