135 public catches this week · all weeks · the live wire

09-19RTNbinary

A new endpoint related to 'auto-mode-classifier-billing' has been detected in the @anthropic-ai/claude-code-darwin-arm64 native binary.

09-19URGENTgithub_releases+pages

Claude Code's auto mode for API and Enterprise users now defaults to a server-side classifier to avoid billing for classifier overhead, with a new status row.

09-18RTNhacker news

Claude Code's changelog indicates it now reads AGENTS.md as a fallback if Claude.md is not present, suggesting a possible shift in documentation or agent-related configurations.

09-18RTNweb app

Anthropic's Claude platform deployed new frontend code, including a new endpoint related to Stripe billing regions.

09-18URGENTbinary

Anthropic's `claude-code-darwin-arm64` package update includes new feature flags related to `tengu` and internal tooling, along with a new model ID and an endpoint: `https://clau.de/wrap-up`.

09-18RTNgithub_releases+pages

Claude Code (v2.1.277) now supports `AGENTS.md` for project instructions and includes various fixes for hanging sessions, error reporting, and UI crashes.

09-18URGENTnpm

Anthropic's SDK update 0.127.0 suggests new context management features, including conversation compaction and handling for tool calls, along with a `myContext` option.

09-18RTNgithub

Anthropic's Python SDK v1.7.0 adds a tool runner compaction feature and new rate limit display names, while deprecating old group_type.

09-18FLASHpypi

Anthropic's 1.7.0 Python SDK update includes references to new models like 'claude-opus-4-8' and 'claude-haiku-4-5', alongside beta features for refusal fallback and environment workers.

09-18FLASHmodel_catalog

191 new model IDs, including new versions of Amazon's Nova, Anthropic's Claude, Google's Gemini, and Cohere's Command models, were added to the LiteLLM model catalog.

09-18RTNgithub_releases+pages

Anthropic's claude-code released v2.1.276, fixing a regression where requests failed with a '400 Input tag 'advisor_20260301'' error when using a proxy or gateway.

09-18RTNbinary

New strings in Anthropic's claude-code binary suggest features around internal policy controls, registry URLs, and experimental loaders.

09-17URGENTgithub_releases+npm+pages

Claude Code's v2.1.275 update adds features like skill/plugin syncing from claude.ai, a send-now key, and improved plugin installation, alongside numerous bug fixes.

09-17RTNweb app

Anthropic's Claude platform console deployed new frontend code including new Stripe billing endpoints and a direct link to claude.ai.

09-17URGENTgithub_org

Anthropic released a new public GitHub repository called "uplifting-biomolecular-modeling."

09-17URGENTbinary

Anthropic's Claude Code for macOS (arm64) gained 18 new strings including a new model_id, feature flags like 'experimentalH' and 'internalModelFieldsRefused', and a '/v1/claude/test/' endpoint. ✓ beat the announcement by 17.3h

09-17RTNgithub_releases+npm+pages

Claude Code (v2.1.274) adds memory usage warnings, OpenTelemetry attributes for LLM requests, and various bug fixes for sessions and tool calls.

09-16RTNhacker news

Anthropic has merged Claude Cowork into the main Claude chat interface, simplifying the user experience.

09-16RTNsitemap

Anthropic added 34 new localized documentation pages to its sitemap, including guides for commerce agents and CLI tools for `apply` and `sessions-connect`.

09-16URGENTbinary

Anthropic's `claude-code` binary shows new internal strings for a potential 'fast retry' feature, model documentation, and billing overage, along with new API endpoints. ✓ beat the announcement by 21.2h

09-15FLASHmodel_catalog

72 new model IDs were added to the litellm catalog, including various unannounced versions of models from OpenAI, Anthropic, Google, and others. The new model IDs include 'gpt-5.3-codex', 'claude-opus-5', 'gemini-3.7-flash', and 'grok-4.6'.

09-15RTNgithub_releases+pages

Anthropic's Claude Code v2.1.273 release introduces new request headers for LLM gateways, session forking, and numerous bug fixes and improvements.

09-15URGENTnpm

Anthropic's SDK update includes a 'myContext' feature flag, along with several JSON examples for tool use, message structures, and continuation summaries, hinting at enhanced context handling and structured interactions. ✓ beat the announcement by 27.9h

09-15RTNweb app

Anthropic's Claude platform console deployed new frontend strings, including an 'internal_tier' feature flag and several new API endpoints, some related to hCaptcha.

09-15URGENTgithub

Anthropic's Python SDK v1.6.0 introduces new API features for Managed Agents, compaction, workspace data residency, input transformations, and web fetch tools, alongside client improvements. ✓ beat the announcement by 31.0h

09-15FLASHpypi

Anthropic's Python SDK v1.6.0 introduces new model IDs, beta features for messages, environments, and tools, and expands on streaming capabilities. ✓ beat the announcement by 31.0h

09-15RTNweb app

Anthropic's Claude platform console deployed new frontend strings referencing workspace management and archiving endpoints, suggesting upcoming features for organizing work.

09-15URGENTbinary

Anthropic's Claude Code binary for Darwin ARM64 was updated, revealing a new model ID `claude-opus-4-5-20251101` and several internal feature flags related to worker behavior, beta features, and tool use. ✓ beat the announcement by 45.2h

09-15URGENTgithub_org

Anthropic released a new public GitHub repository: 'claude-for-financial-advisors', indicating a Claude Cowork plugin with workflows and third-party connectors for financial advisors. ✓ beat the announcement by 45.4h

09-15RTNgithub_releases+pages

Anthropic's claude-code repository released version v2.1.272, primarily focused on bug fixes and reliability improvements.

09-14URGENThacker news

An HN story is gaining rapid traction, alleging a single firm is responsible for hacking scandals at OpenAI, Anthropic, and Meta.

09-14URGENTgithub_releases+npm+pages

Claude Code gains a 'fast mode' in remote sessions, mouse support for config, and granular `allowed_domains` for sandboxed commands, alongside numerous bug fixes and a bump to v2.1.271. ✓ beat the announcement by 48.0h

09-14RTNhacker news

An article titled "Claude is a Contrarian" is gaining rapid traction on Hacker News, indicating community interest in Claude's behavior or performance.

09-14RTNweb app

Anthropic's Claude platform added new API endpoints related to code sessions, environment providers, and Stripe billing regions, alongside hCaptcha assets.

09-14URGENThacker news

High velocity HN story suggests Apple's Siri AI could be swapped for third-party models like Claude or ChatGPT, according to code analysis.

09-20URGENTgithub

OpenAI's Codex project has released a new alpha version, rust-v0.156.0-alpha.9, indicating ongoing development in its Rust implementation.

09-19URGENTnpm

OpenAI's latest SDK hints at new 'safety cases' API, 'ranker' options, enhanced 'file_search' tools with chunking controls, and dedicated JSON mode instructions.

09-19RTNgithub

OpenAI's Codex project has released a new alpha version, rust-v0.156.0-alpha.8.

09-19URGENThacker news

An HN post titled "GPT-6 Astra Solves a WWI German Radio Cipher" is gaining traction, suggesting a potential new model or capability. ✓ beat the announcement by 26.9h

09-19RTNgithub

openai/codex released a new alpha version, `rust-v0.156.0-alpha.7`.

09-19RTNgithub

A new pre-release version rust-v0.156.0-alpha.6 has been published for openai/codex.

09-19URGENTgithub

OpenAI's Codex project has published a new pre-release, `rust-v0.156.0-alpha.5`, indicating active development.

09-18FLASHsitemap

OpenAI quietly published 225 new sitemap URLs, all related to various business plugins, suggesting an expansion of their plugin ecosystem for business use cases.

09-18RTNgithub

A new pre-release, rust-v0.156.0-alpha.4, has been published for openai/codex.

09-18RTNgithub

OpenAI Python library v3.16.2 fixes a memory leak in response parsing.

09-18FLASHpypi

OpenAI's 3.16.2 update includes a new model ID 'gpt-4o-2024-08-06', potential 'realtime' and 'responses' beta features, and updates to JSON mode and streaming APIs.

09-18RTNnpm

The OpenAI SDK v7.19.0 includes updates for file search tool controls like chunking and overlap, a new ranker 'default_2024_08_21', and clearer documentation around JSON mode.

09-18RTNgithub

OpenAI's Python library released v3.16.1, which fixes an issue with loading unrelated API resources on first use.

09-18FLASHpypi

OpenAI's latest SDK update, version 3.16.1, hints at a new `gpt-4o-2024-08-06` model, adds 'beta' connection features for 'realtime' and 'responses', and introduces 'live', 'forks', and 'sideband' connections. It also improves JSON mode and streaming APIs.

09-18RTNgithub

OpenAI Codex published a pre-release for its Rust library, version 0.156.0-alpha.3.

09-18RTNgithub

OpenAI's Python library adds webhook endpoint management and deprecates MCP connector_id in v3.16.0.

09-18FLASHpypi

OpenAI's latest client library update hints at a new `gpt-4o-2024-08-06` model, new 'realtime' and 'responses' beta connections, and a lot of webhook endpoint management features.

09-18RTNgithub

A new pre-release version, rust-v0.156.0-alpha.2, has been published for openai/codex.

09-18RTNhacker news

A blog post titled "Hacking OpenAI" is gaining traction on Hacker News, indicating community interest in potential vulnerabilities or exploits related to OpenAI systems.

09-18RTNgithub

OpenAI Codex published a new pre-release, rust-v0.155.0-alpha.9.2, suggesting ongoing development in the Rust client.

09-18FLASHpypi

OpenAI's latest `openai` package (3.15.0) includes references to a new `gpt-4o-2024-08-06` model and introduces `beta.realtime` and `beta.responses` WebSocket connections, possibly for streaming or real-time interaction.

09-18RTNgithub

OpenAI's Codex repository published a new Rust pre-release, version `rust-v0.156.0-alpha.1`.

09-18URGENTgithub

OpenAI's Python library adds an 'audio-mini' model choice, agent session settings, prompt cache prewarming, and managed WebSocket sessions.

09-18URGENTnpm

OpenAI's latest SDK update (7.18.0) hints at new `file_search` tool features like `chunk_overlap_tokens` and `chunk_size_tokens`, potentially improving search relevance. It also introduces a new `ranker` option with a `default_2024_08_21` value, and mentions `JSON mode` which forces JSON responses.

09-17URGENTgithub

OpenAI's Codex project published a new pre-release version: rust-v0.155.0-alpha.18.

09-17URGENTgithub

OpenAI's Codex Rust v0.155.0 release includes experimental voice conversations, live reasoning summaries in the TUI, enhanced agent management, and Amazon Bedrock credential improvements.

09-17URGENThacker news

An OpenAI-related link titled "Astra for Law" is climbing quickly on Hacker News, suggesting a potential new legal-focused AI offering. ✓ beat the announcement by 32.0h

09-17URGENTgithub

OpenAI's Codex project has released a new pre-release version: rust-v0.155.0-alpha.17.

09-17RTNweb app

A new staging endpoint `chatgpt-staging.com` was found in a ChatGPT web app frontend deploy, indicating possible internal testing or upcoming features.

09-17FLASHweb app

OpenAI's ChatGPT web app deployed a new frontend bundle, revealing strings for 'gpt-5' and various 'gpt-4o-gdrive' and 'gpt-5-x-reasoning' model IDs, alongside internal feature flags and new endpoints.

09-17RTNweb app

ChatGPT web app adds a new staging endpoint, potentially indicating upcoming features or testing.

09-17FLASHweb app

OpenAI's ChatGPT web app has new strings referencing 'gpt-5' and various 'gpt-5' sub-models, along with 'gpt-4o-mini' and 'o4-mini'. There are also many 'internal' feature flags and new endpoints, including some for Google Drive integration.

09-17RTNgithub

OpenAI's Codex project released a new pre-release version, rust-v0.155.0-alpha.16, indicating ongoing development, possibly with a focus on Rust integration or improvements.

09-17RTNsitemap

OpenAI quietly added 10 new business plugin pages to its sitemap, all focused on legal and compliance applications.

09-17RTNgithub

OpenAI Codex published a new pre-release: rust-v0.155.0-alpha.15, suggesting ongoing development in the Rust integration for Codex.

09-16RTNgithub

OpenAI's Codex project published a new pre-release, rust-v0.155.0-alpha.14, suggesting ongoing development in the Rust implementation.

09-16RTNgithub

OpenAI's Codex project released a new pre-release version: rust-v0.155.0-alpha.13.

09-16FLASHmodel_catalog

LiteLLM model catalog added 22 new model IDs including several `gpt-5.5`, `gpt-5.6`, `gpt-6`, and `gpt-image` variants on Azure, plus new Gemini, Command, Mistral, and other provider models. ✓ beat the announcement by 22.1h

09-16URGENTgithub

OpenAI's Codex project published a new pre-release version: rust-v0.155.0-alpha.2.6. ✓ beat the announcement by 24.0h

09-16URGENTnpm

OpenAI's latest SDK update introduces 'ranker' parameter for Assistants, adds new 'code_interpreter' and 'file_search' tool types with detailed file search configurations like chunking and overlap, and clarifies JSON mode usage.

09-16RTNgithub

OpenAI Codex published a new pre-release for its Rust SDK, version 0.155.0-alpha.12.

09-16FLASHmodel_catalog

59 new model IDs, including mentions of GPT-4.1, GPT-5, GPT-5.x, GPT-6, and O1/O3/O4 models, appeared in the LiteLLM model catalog for Azure EU and US regions. ✓ beat the announcement by 0.9h

09-16URGENTgithub

OpenAI's Codex project published a new pre-release: `rust-v0.155.0-alpha.11`. ✓ beat the announcement by 2.0h

09-16URGENTnpm

OpenAI's latest SDK hints at new 'file_search' tool overrides with chunking controls, a 'ranker' option, and expanded JSON mode guidance.

09-16RTNsitemap

OpenAI quietly published 9 new pages in their sitemap related to disrupting malicious uses of AI, specifically focusing on cyber special operations, date bait, false witness, fish food, no bell, romance scams, silver lining playbook, and trolling stone.

09-16URGENThacker news

OpenAI is expanding ChatGPT ads with new 'Sponsored Agents', suggesting a move towards monetizing the platform through advertising partnerships. ✓ beat the announcement by 4.0h

09-16URGENTgithub

OpenAI's Codex project published a new pre-release for its Rust-based version, `rust-v0.155.0-alpha.2.5`. ✓ beat the announcement by 5.9h

09-16URGENTgithub

OpenAI's Codex project published a pre-release version named `rusty-v8-v152.2.0`. ✓ beat the announcement by 5.9h

09-16URGENTgithub_org

OpenAI made a new public repository called 'community-plugins', suggesting a potential platform for community-contributed plugins.

09-16RTNgithub

A new pre-release, rust-v0.155.0-alpha.10, has been published for openai/codex.

09-16URGENTgithub

OpenAI's Codex project has published a new pre-release, `rust-v0.155.0-alpha.9`. ✓ beat the announcement by 16.8h

09-16FLASHpypi

OpenAI's latest SDK update hints at a new `gpt-4o-2024-08-06` model and introduces `beta.realtime` and `beta.responses` connection features.

09-15RTNgithub

OpenAI's Codex project has released a new pre-release version: rust-v0.155.0-alpha.8.

09-15RTNgithub

OpenAI's python library (v3.14.1) includes several bug fixes and documentation updates, plus a chore that bumps an example to `gpt-5.6`.

09-15RTNgithub

OpenAI Codex published a new pre-release for its Rust library, version 0.155.0-alpha.7.

09-15URGENThacker news

Hugging Face is reportedly billing OpenAI $100M for compute costs related to alleged unauthorized use of their platform.

09-15RTNgithub

OpenAI's Codex project released a new pre-release version: rust-v0.155.0-alpha.6, indicating ongoing development in the Rust implementation.

09-15RTNgithub

OpenAI's Codex repository published a pre-release for rust-v0.155.0-alpha.5.

09-15RTNgithub

OpenAI Python library v3.14.0 improves error handling for streaming and fixes several bugs, including vector store file polling and API error code normalization.

09-15FLASHpypi

OpenAI's latest `openai` package (3.14.0) hints at a new `gpt-4o-2024-08-06` model and introduces `beta.realtime` and `beta.responses` connections, possibly for real-time interaction or streaming responses.

09-14RTNgithub

OpenAI's Codex project published a new pre-release for its Rust-based version, 0.155.0-alpha.2.4, on GitHub.

09-14FLASHmodel_catalog

34 new model IDs were added to the LiteLLM model catalog, including several Qwen, DeepSeek, Llama, and Gemini models, with a notable 'gpt-5.5-cyber' and 'gemini-robotics-er-2' appearing. ✓ beat the announcement by 44.0h

09-14FLASHlmarena

A new model, hidream-o1-video-vcyz, has appeared on the LMArena leaderboard, suggesting it's an unreleased model currently in testing.

09-14URGENThacker news

An HN post discussing "GPT-5.6 Luna" and "GPT-6 Astra" is gaining traction, suggesting new model names or performance comparisons are being explored. ✓ beat the announcement by 46.0h

09-14FLASHmodel_catalog

Three new Azure model IDs, including 'chat-latest' and 'gpt-chat-latest', were added to the LiteLLM model catalog.

09-14RTNweb app

ChatGPT web app updated frontend code, adding new endpoint strings for chat.openai.com, chatgpt-staging.com, and chatgpt.com.

09-14RTNgithub

OpenAI's Codex project released a new pre-release version: rust-v0.155.0-alpha.4.

09-19RTNgithub

Gemini CLI published a new pre-release v0.62.0-nightly.20260919.gcfbcaa8df with fixes for ConPTY process exit, request cancellation logs, terminal memory, Windows paths, VSCode terminal focus, and auth error documentation.

09-18RTNgithub

Google Gemini CLI published a new pre-release `v0.62.0-nightly.20260918.g9450ade79` with fixes for OAuth token retention, UI layout, PTY file descriptor cleanup, and integration test flakiness.

09-17RTNgithub

A new nightly pre-release, v0.62.0-nightly.20260917.g6a466a7e2, has been published for the Google Gemini CLI.

09-16RTNgithub

Gemini CLI published a new nightly pre-release v0.62.0, with fixes to `AgentLoopContext` properties and `MCP tool call` formatting.

09-15RTNgithub

Google Gemini CLI released a v0.61.0-preview.0 with fixes for Flash model IDs and security hardening, alongside a v0.60.0 stable release focused on security improvements.

09-15FLASHhacker news

An HN post titled "Gemini 3.8 Live and 3.8 Live Extended Thinking" is gaining rapid traction with 44 pts/hr.

09-15RTNgithub

A new nightly pre-release, v0.61.0-nightly.20260915.g9c1b0a610, was published for the Google Gemini CLI, indicating ongoing development.

09-14FLASHmodel api

A new model ID, "antigravity-preview-09-2026," has appeared in Google Gemini's API, suggesting a potential future release or preview.

09-19FLASHmodel_catalog

Two new xai/grok-voice-transcribe model IDs (1.0 and 2.0) detected in the LiteLLM model catalog, suggesting new voice transcription capabilities.

09-18FLASHmodel api

OpenRouter's API added a new model ID: deepseek/deepseek-v4-flash-0731:free.

09-17FLASHhuggingface

Nvidia has published a new HuggingFace repository for 'DeepSeek-V4.1-Flash-NVFP4'.

09-16RTNhacker news

Mistral AI and Mozilla are collaborating on private, multilingual AI browsing, as indicated by a fast-climbing Hacker News story.

09-20FLASHhuggingface

Qwen pushed new HuggingFace repositories for Qwen-Image-2.1, including variants for image-to-image (I2I) and text-to-image (T2I).

09-18FLASHmodel_catalog

LiteLLM's model catalog now includes four new Qwen 3.8 models from DashScope and Qwen AI Platform, including flash and omni-flash versions.

09-17FLASHmodel api

A new model ID, qwen/qwen3.8-27b:free, has appeared in OpenRouter's API, suggesting a potential upcoming release.

09-16RTNsitemap

Cursor quietly published two new sitemap pages related to "federal" and "federal/compliance," indicating a potential expansion into government or regulated sectors.

09-15RTNdocs

Cursor.com updated its changelog to include 'Federal' and 'Compliance' in relation to its Enterprise offerings, suggesting new tiering or features.

Six new ByteDance Seedance 2.0 and 2.5 image, reference, and text-to-video models added to the LiteLLM model catalog via fal_ai.

A new service called Pirate Face is gaining traction for its ability to rescue LLM models from deletion.

LiteLLM model catalog added two new models: openrouter/z-ai/glm-5.3-flashx and openrouter/prism-ml/ternary-bonsai-2-27b.

OpenRouter API now lists a new model ID: z-ai/glm-5.3-flashx, likely indicating an upcoming release.

Nvidia created a new HuggingFace repository: nvidia/PixelUMM-v1, possibly indicating a new unannounced model.

Litellm model catalog added 7 new model IDs, including Azure_AI/FLUX.2-flex and several Deepgram streaming models for Nova-3, multilingual, and various audio features like entity detection and diarization.

New model ID transcribe/StartTranscriptionJob detected in LiteLLM's model catalog.

Nvidia created a new HuggingFace repository named 'nvidia/cari4d_commercial'.

Three new model IDs (typesafe/jev-1.13.0, typesafe/jev-latest, typesafe/jev-preview) have been added to the model catalog.

Nvidia created a new HuggingFace repository named 'nvidia/GLM-5.3-NVFP4'.

A new model ID, 'wandb/zai-org/GLM-5.3-Flash', has been added to the LiteLLM model catalog.

A Hacker News article titled "I Don't Like LLMs" by Martin Fowler is rapidly gaining traction, reaching 70 points in 1.4 hours (50 pts/hr).

LiteLLM's model catalog now lists 'openrouter/stealth/union-alpha', suggesting a new model from OpenRouter.

A paper titled "Breaking the 1.58-bit Barrier for Ternary LLMs" is rapidly gaining traction on Hacker News.

The PS5 Linux lead has quit, citing concerns about "noobs using LLMs" in open-source projects without understanding them.

Nvidia created two new HuggingFace repositories: `nvidia/c-foundationstereo-s` and `nvidia/foundationpose`. These could be early indicators of new models or datasets.