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

07-31FLASHlmarena

Two new models, claude-opus-5-max-image-to-webdev and kimi-k3-webdev-image-to-webdev, have appeared on the LMArena leaderboard, suggesting upcoming image-to-web development capabilities.

07-29FLASHhacker news

Claude's service is experiencing an outage, evidenced by a fast-climbing Hacker News story.

07-29FLASHgithub

OpenAI's rust-v0.146.0 release for Codex introduces new session management, expanded plugin marketplace support including Amazon Bedrock and Claude Code, thread forking, web search for custom models, and executor-provided skill discovery. ✓ beat the announcement by 15.1h

07-28URGENThacker news

An Anthropic research paper on "Discovering Cryptographic Weaknesses with Claude" is gaining rapid traction on Hacker News, suggesting Claude's capabilities in security analysis.

07-28FLASHpypi

Anthropic's latest SDK update, 0.120.2, includes references to new models 'claude-opus-4-8' and 'claude-haiku-4-5', alongside 'BetaRefusalFallbackMiddleware' indicating a new beta fallback mechanism for messages.

07-28FLASHpypi

Anthropic's latest SDK update (0.120.1) reveals new model IDs "claude-opus-4-8" and "claude-haiku-4-5" along with beta streaming and refusal fallback middleware features.

07-28FLASHmodel api

OpenRouter API shows new batch inference models, including numerous Claude 4.x/5, Gemini 2.5/3.x, and GPT-5.x variants, suggesting unannounced future models from Anthropic, Google, and OpenAI.

07-28URGENThacker news

Anthropic's "Our position on open-weights models" is gaining significant traction on Hacker News, indicating a notable public interest in their stance on open-source AI.

07-27FLASHlmarena

Two new Claude Opus 5 models, 'claude-opus-5-max-text' and 'claude-opus-5-max-webdev', have appeared on the LMArena leaderboard, suggesting potential unreleased versions.

07-27FLASHlmarena

LMArena has added 8 new models to its leaderboard, including several 'claude-opus-5-high' variants (document, text, vision, webdev), 'grok-4.5' (document, vision), and 'super-nova-ext-3tam' (document, vision).

08-02FLASHweb app

ChatGPT web app frontend deploy [fffaf0c8]: 371 new strings (187 endpoint, 138 feature_flag, 48 model_id, 16 sentence) — verify

08-02RTNweb app

ChatGPT web app frontend deploy [bfe3fd5d]: 12 new strings (12 endpoint) — verify

08-01URGENTgithub_org

OpenAI released a new public GitHub repository called 'ten-proofs' focused on Lean certificates for proofs in mathematics and theoretical computer science.

08-01RTNweb app

ChatGPT web app update includes new feature flags, potentially hinting at image generation composer CTA and internal conversation rendering.

07-31RTNweb app

New feature flags detected in ChatGPT web app bundle, potentially related to prompt textarea focus and internal content rendering.

07-31FLASHlmarena

LMArena has listed five new models, including several gpt-5.6 variants, grok-4.5-image-to-webdev, and super-nova-ext-3tam-image-to-webdev, hinting at advanced image-to-web development capabilities.

07-31URGENTsitemap

OpenAI added 44 new pages to its sitemap, all related to "disrupting malicious uses of AI" with various specific threat actor and scam names.

07-31RTNgithub

OpenAI's Codex project has published a new pre-release, `rust-v0.147.0-alpha.4`.

07-31URGENTgithub

OpenAI's Codex project released a new pre-release version: rust-v0.147.0-alpha.3.

07-31RTNgithub

OpenAI Python SDK v2.52.0 adds content provenance checks, honors Retry-After delays up to two minutes, and includes API-key mTLS HTTP client recipes.

07-31FLASHpypi

OpenAI's latest SDK update (2.52.0) hints at a new `gpt-4o-2024-08-06` model, and introduces 'realtime' and 'responses' beta connections, alongside new provenance checking features.

07-31URGENTnpm

OpenAI's 7.3.0 update introduces a new `ranker` option with `default_2024_08_21`, file search tool overrides including `chunk_overlap_tokens` and `chunk_size_tokens`, and clarifies JSON mode usage.

07-31FLASHweb app

OpenAI's ChatGPT web app has new strings indicating potential GPT-5 integration, early access to experimental features, and health-related data integration, alongside various payment and analytics updates.

07-31URGENTgithub

OpenAI's Codex project has released a new alpha version, rust-v0.147.0-alpha.1.1, indicating continued development and potential upcoming features for its Rust-based components.

07-31RTNweb app

ChatGPT web app updated with new strings, including several external links and internal feature flags related to content pages and conversation rendering.

07-31FLASHlmarena

10 new gpt-5.6 models with 'luna', 'sol', and 'terra' prefixes and specialized 'xhigh' capabilities (code, document, text, vision) appeared on the LMArena leaderboard.

07-30RTNweb app

ChatGPT web app update includes new feature flags related to 'automations', 'focusPromptTextarea', and various 'contentful' and 'internal-render-conversation' assets.

07-30RTNweb app

ChatGPT web app update includes 8 new strings, potentially related to internal features like file translation, conversation rendering, and automation scheduling, along with an "icebreaker-dogfooding" flag.

07-30RTNgithub

OpenAI's Python library adds a new 'fast tier' to its API, potentially indicating new performance options.

07-30FLASHpypi

OpenAI's latest `openai` package (2.51.0) includes a new model ID `gpt-4o-2024-08-06` and introduces beta `realtime` and `responses` connection features, suggesting upcoming streaming or interactive capabilities beyond standard completions.

07-30FLASHweb app

OpenAI's ChatGPT web app has new strings indicating internal testing of multiple new GPT models, including 'GPT-4.1', 'GPT-4.1-mini', 'GPT-4o', 'GPT-5', 'GPT-5-mini', 'GPT-5-pro', and 'GPT-5 Thinking Mini', alongside new GPT-4 usage cap messages.

07-30FLASHhn+sitemap

OpenAI's sitemap and an HN story indicate the quiet publication of pages related to a 'GPT-5.6' model, API fast mode, and new business partners.

07-30FLASHdocs

OpenAI is introducing new GPT-5.6 models (sol, terra, luna) and replacing 'Priority Processing' with 'Fast mode' in the API, along with price reductions for some models.

07-30RTNnpm

OpenAI's latest SDK update hints at a new 'ranker' option ('default_2024_08_21'), detailed file search chunking controls, and clarifies JSON mode usage.

07-30RTNgithub

A new pre-release, rust-v0.146.0-alpha.9.2, has been published for openai/codex, indicating ongoing development in the Rust implementation.

07-30RTNgithub

OpenAI's Codex project published a new pre-release version: rust-v0.147.0-alpha.2.

07-30RTNweb app

New internal feature flags detected in ChatGPT web app bundle, potentially related to automation, scheduling, or internal content rendering.

07-29RTNgithub

OpenAI's Codex project has published a new pre-release version: rust-v0.146.0-alpha.9.1.

07-29FLASHweb app

OpenAI's ChatGPT web app has new strings referencing several unreleased GPT models, including GPT-4.1, GPT-4.5, GPT-5, GPT-5.1, GPT-5.2, GPT-5.3, and various 'mini', 'pro', and 'reasoning' versions, alongside internal feature flags and endpoints.

07-29URGENTdocs

OpenAI released an official Terraform provider for managing API Platform resources as infrastructure as code.

07-29RTNweb app

ChatGPT web app frontend deploy included 4 new feature flag strings, potentially indicating upcoming features or internal tooling.

07-29RTNweb app

ChatGPT web app frontend deploy includes new feature flags related to dynamic content and internal conversation rendering.

07-29RTNweb app

New strings in the ChatGPT web app frontend suggest internal feature flags, contentful dynamic feature pages, and possible conversation rendering updates.

07-29URGENTgithub

OpenAI's Codex project has released a new alpha version, rust-v0.147.0-alpha.1, indicating ongoing development, specifically with a Rust focus. ✓ beat the announcement by 10.5h

07-29RTNhacker news

An OpenAI-related 'Codex Security' project is gaining significant traction on Hacker News.

07-29URGENTgithub

OpenAI Codex published a pre-release version `rusty-v8-v150.4.0`, indicating potential updates related to V8 engine integration. ✓ beat the announcement by 15.1h

07-29RTNgithub

OpenAI Python library v2.50.0 includes transcription model updates and a fix for a transcription keyword overload.

07-29FLASHpypi

OpenAI's latest client library (v2.50.0) introduces a new `gpt-4o-2024-08-06` model ID and beta features for "realtime" and "responses" connections, hinting at real-time streaming capabilities.

07-29URGENTnpm

OpenAI's latest SDK update (7.1.0) includes new configuration options for 'file_search' tools, with settings for 'chunk_overlap_tokens' and 'chunk_size_tokens', alongside a new 'ranker' option, potentially for model response ranking.

07-29FLASHweb app

OpenAI's ChatGPT web app added strings for 'GPT-5' and several new features like image generation prompt sharing, live translation, read aloud, and transcription.

07-28FLASHdocs

OpenAI added gpt-transcribe and gpt-live-transcribe models for file and real-time audio transcription, respectively, with new features like context, hints, and multi-language support.

07-28FLASHweb app

ChatGPT web app update includes references to "GPT-5", "early access to experimental features," and various internal health-related testing flags, potentially hinting at future model updates and testing for medical data integration.

07-28FLASHweb app

OpenAI's ChatGPT web app update shows numerous new model IDs including "GPT-4.1", "GPT-4.1-mini", and various "GPT-5" iterations like "GPT-5 Thinking Mini", "GPT-5-mini", "GPT-5.2", "GPT-5.3", "GPT-5-pro", and "GPT-5-reasoning".

07-28RTNdocs

OpenAI's changelog now references an 'llms.txt' file, potentially indicating a new documentation index or internal listing of language models.

07-28URGENTgithub_org

OpenAI made a new public repository called 'codex-security' which includes SDKs and CLI for Codex Security.

07-28RTNgithub

A new pre-release, rust-v0.146.0-alpha.14, was published for openai/codex.

07-28FLASHpypi

OpenAI's latest SDK update (2.49.0) introduces `gpt-4o-2024-08-06`, and new beta features for `realtime` and `responses` connections, alongside various client-side improvements and validator messages for fine-tuning.

07-28FLASHnpm

OpenAI's 7.0.0 npm package hints at new 'file_search' and 'code_interpreter' tool types, 'ranker' options, and 'JSON mode' for models.

07-28RTNweb app

ChatGPT web app update includes new feature flags related to dynamic content and internal conversation rendering, possibly indicating upcoming content features or backend improvements.

07-27RTNgithub_org

OpenAI made a new public repository called 'birdingpal' which is a ChatGPT-enabled tool for bird identification.

07-27RTNweb app

New feature flags detected in the ChatGPT web app frontend deploy. This could indicate upcoming changes or features related to content pages and internal conversation rendering.

07-27RTNgithub

OpenAI's Codex project published a new pre-release, rust-v0.146.0-alpha.13, suggesting ongoing development in its Rust implementation.

07-27RTNgithub

A new pre-release, rust-v0.146.0-alpha.12, has been published for openai/codex.

07-27RTNweb app

ChatGPT web app updated with 5 new strings, likely feature flags, including references to internal content rendering and contentful dynamic feature pages.

08-02URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.55.0-nightly.20260802.gf47d6c6f7

08-01RTNgithub

Gemini CLI published a new pre-release `v0.55.0-nightly.20260801.gf47d6c6f7` with fixes for capacity exhaustion retries and improved error details for empty stream responses.

07-31RTNgithub

Gemini CLI published a new pre-release, v0.54.0-preview.1, indicating ongoing development, alongside a patch release v0.53.1.

07-30FLASHhacker news

A Google DeepMind blog post about "Gemini Robotics 2" is gaining significant traction on Hacker News, suggesting an imminent or recent announcement related to whole-body intelligence for robots.

07-30FLASHmodel api

A new model ID, 'gemini-robotics-er-2-preview', has appeared in Google Gemini's API, indicating a potential robotics-focused model.

07-30RTNgithub

Google Gemini CLI published a new pre-release, v0.55.0-nightly.20260730.gdc859e8e4, indicating ongoing development and future updates.

07-29URGENTgithub

Gemini CLI published two new pre-releases (v0.55.0-nightly, v0.54.0-preview) and a stable release (v0.53.0), introducing new features for PR generation, agent runners, and several core fixes.

07-28URGENTgithub_org

Google Gemini made a new public repository called 'glanceboard'.

07-28RTNgithub

Gemini CLI published a new pre-release `v0.54.0-nightly.20260728.gbef611950` with fixes for line endings in `getProposedContent` and tag length validation in the file keychain.

07-27RTNgithub

A new nightly pre-release, v0.54.0-nightly.20260727.g3818efbbf, for the Google Gemini CLI was published.

07-28URGENTdocs

Cursor is launching a new 'Start' plan in India for ₹649/month, offering access to Grok 4.5 and Composer models, cloud agents, iOS app remote control, plugins, and local UPI payments.

07-27RTNdocs

Cursor's changelog updated to include 'Models', 'Grok', 'Composer', and 'Evals' as new categories or mentions.

08-01FLASHlmarena

A new model, deepseek-v4-flash-high-20260730-webdev, appeared on the LMArena leaderboard, suggesting potential future release.

07-31FLASHlmarena

DeepSeek has added two new models, deepseek-v4-ch3-thinking-webdev and deepseek-v4-flash-thinking-20260730, to the LMArena leaderboard, suggesting upcoming releases.

07-31RTNhacker news

An analysis of DeepSeek V4 Flash 0731's intelligence, performance, and price is gaining traction on Hacker News.

07-31FLASHhuggingface

deepseek-ai pushed a new HuggingFace repo for what appears to be a new model named DeepSeek-V4-Flash-0731.

07-31FLASHmodel api

A new DeepSeek model, deepseek/deepseek-v4-flash-0731, has appeared in OpenRouter's API, suggesting an upcoming release.

07-31RTNhacker news

DeepSeek-V4-Flash Update is trending on Hacker News, indicating a recent announcement or significant activity from DeepSeek.

07-30FLASHmodel_catalog

LiteLLM's model catalog now includes two new model IDs: dashscope/qwen3.7-max and dashscope/qwen3.7-plus, suggesting new models from Alibaba Cloud's large model family.

07-28FLASHmodel api

New model id `qwen/qwen3.7-flash` spotted in OpenRouter's API, suggesting a new Qwen model release.

08-01RTNhacker news

HN velocity: "Cursor removed cost information from the usage page and CSV export" (44 pts/hr)

07-29URGENTdocs

Cursor is now available on iPad for paid plans, with new features like an inbox and a full PR review experience on both iPhone and iPad.

A new anonymous model, inkling-small-rc-3-2ud8-text, has appeared on the LMArena leaderboard, suggesting a stealth test of a small release candidate model.

A fast-moving HN story details how 'AI worms' could self-propagate through Microsoft Copilot for Word.

Nvidia created a new HuggingFace repository named 'nvidia/stormcast-conus'.

Moonshot AI created a new HuggingFace repository named 'moonshotai/Kimi-K3'. This could potentially be a new model, dataset, or space related to their 'Kimi' series.

LMArena has added a new anonymous model, "hy3-tencent-cloud-text," to its leaderboard, likely a stealth model from Tencent Cloud.