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

08-16RTNhacker news

HN velocity: "Claude: System Prompts" (65 pts/hr)

08-16URGENTgithub_org

anthropics made 2 repos public: amulet2, mockturtle

08-15FLASHgithub_releases+pages

Claude Code changelog changelog/docs page changed (23 new lines) — corroborated across 2 sources (github_releases, pages)

08-15RTNweb app

Anthropic Console frontend deploy [aca8e3c3]: 2 new strings (2 feature_flag, 2 sentence) — verify

08-15URGENTbinary

@anthropic-ai/claude-code-darwin-arm64 2.1.233 native binary: 30 new strings (19 feature_flag, 12 sentence, 11 endpoint) — verify

08-14URGENTweb app

Anthropic's Claude console received a frontend update with new strings hinting at detailed output token breakdown and internal reasoning token counts, alongside new API endpoints for user access and KYC status.

08-14URGENTbinary

@anthropic-ai/claude-code-darwin-arm64 2.1.232 native binary: 67 new strings (34 endpoint, 30 sentence, 24 feature_flag, 9 model_id) — verify

08-13FLASHgithub_releases+pages

anthropics/claude-code published a release: v2.1.232 — corroborated across 2 sources (github_releases, pages)

08-13URGENTnpm

@anthropic-ai/sdk 0.117.1 shipped: 35 interesting new strings (35 sentence, 1 feature_flag)

08-13RTNappstore

Claude (iOS) app updated to 1.260813.0 — release notes: Squashed some bugs and improved the overall experience. Yours, Claude

08-13RTNgithub

anthropics/anthropic-sdk-python published a release: v0.122.0

08-13FLASHpypi

anthropic 0.122.0 shipped: 41 interesting new strings (41 sentence, 2 model_id, 2 feature_flag)

08-13RTNsitemap

Anthropic docs sitemap: 11 new pages appeared in the sitemap (unlinked/quietly published)

08-13RTNweb app

Anthropic Console frontend deploy [01b7297a]: 4 new strings (2 feature_flag, 2 sentence, 2 endpoint) — verify

08-13URGENTbinary

@anthropic-ai/claude-code-darwin-arm64 2.1.231 native binary: 301 new strings (281 feature_flag, 105 sentence, 21 endpoint) — verify

08-13RTNhacker news

HN velocity: "If I own Claude's outputs why can't I train my own model on them?" (52 pts/hr)

08-13FLASHgithub_releases+pages

anthropics/claude-code published a release: v2.1.231 — corroborated across 2 sources (github_releases, pages)

08-13URGENTct_logs

anthropic.com: 42 new subdomains appeared in certificate-transparency logs

08-12FLASHgithub_releases+pages

anthropics/claude-code published a release: v2.1.229 — corroborated across 2 sources (github_releases, pages)

08-12RTNweb app

Anthropic Console frontend deploy [36e8787e]: 2 new strings (2 feature_flag, 2 sentence) — verify

08-12URGENTbinary

@anthropic-ai/claude-code-darwin-arm64 2.1.228 native binary: 41 new strings (27 feature_flag, 17 endpoint, 9 sentence, 1 model_id) — verify

08-12RTNsitemap

Anthropic docs sitemap: 68 new pages appeared in the sitemap (unlinked/quietly published)

08-12FLASHweb app

Anthropic Console frontend deploy [d4be6727]: 874 new strings (859 endpoint, 825 feature_flag, 11 model_id, 6 sentence) — verify

08-12RTNhacker news

HN velocity: "Someone is running mass vulnerability scans, spoofing AI bots like ClaudeBot" (36 pts/hr)

08-12RTNappstore

Claude (iOS) app updated to 1.260810.0 — release notes: Squashed some bugs and improved the overall experience. Yours, Claude

08-12FLASHmodel_catalog

15 new model IDs detected, including several new Claude, Gemini, Mistral, and XAI Grok models, with a focus on robotics, moderation, and real-time audio features.

08-11FLASHgithub_releases+pages

Claude Code changelog changelog/docs page changed (19 new lines) — corroborated across 2 sources (github_releases, pages)

08-11URGENTdocs

Anthropic API changelog changelog/docs page changed (96 new lines)

08-11URGENTweb app

Anthropic Console frontend deploy [dec45131]: 19 new strings (16 endpoint, 4 feature_flag, 1 sentence) — verify

08-11FLASHhacker news

Claude Code is reportedly leaking real email addresses via the User-Agent string in curl commands, with a fast-rising Hacker News story.

08-11URGENTbinary

Anthropic's Claude Code for macOS arm64 updates with new internal tooling around session goals, authentication, and error handling, along with new API endpoints related to session management and notifications. It also mentions a new feature called "Routines" which can be enabled by an admin.

08-10RTNnpm+pages

Claude Code changelog shows fixes for feature flag evaluation, Bash command failures, and TUI conversation issues, along with improved slash-command menu and performance.

08-10RTNhacker news

An article discussing how Claude marks AI-generated content is gaining traction on Hacker News, indicating increased interest in AI content provenance.

08-10URGENTweb app

Anthropic's Claude platform console deployed new strings indicating upcoming features like workspace management, organization onboarding, KYC status checks, and a new auto-recharge feature flag.

08-10RTNhacker news

An Anthropic research paper on Claude's mathematical capabilities, specifically regarding the Riemann Zeta function, is gaining significant traction on Hacker News.

08-10URGENTgithub_org

Anthropic made a new public GitHub repo called 'zeta-23-lean'.

08-16FLASHweb app

ChatGPT web app frontend deploy [e1707b42]: 295 new strings (159 endpoint, 119 feature_flag, 18 model_id, 14 sentence) — verify

08-16FLASHweb app

ChatGPT web app frontend deploy [453b5622]: 15 new strings (11 feature_flag, 4 model_id) — verify

08-16URGENTweb app

ChatGPT web app frontend deploy [c8b1009b]: 13 new strings (12 endpoint, 1 feature_flag) — verify

08-16URGENTweb app

ChatGPT web app frontend deploy [54b84c30]: 13 new strings (12 endpoint, 1 feature_flag) — verify

08-16URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.20

08-15FLASHweb app

ChatGPT web app frontend deploy [e1707b42]: 301 new strings (151 endpoint, 129 feature_flag, 22 model_id, 14 sentence) — verify

08-15URGENTweb app

ChatGPT web app frontend deploy [3900e11a]: 13 new strings (12 endpoint, 1 feature_flag) — verify

08-15FLASHweb app

ChatGPT web app frontend deploy [e1707b42]: 15 new strings (11 feature_flag, 3 sentence, 3 endpoint, 1 model_id) — verify

08-15FLASHweb app

ChatGPT web app frontend deploy [025aed95]: 297 new strings (148 endpoint, 128 feature_flag, 22 model_id, 14 sentence) — verify

08-15RTNweb app

ChatGPT web app frontend deploy [ab0c15e5]: 12 new strings (12 endpoint) — verify

08-15URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.19

08-15RTNgithub

openai/openai-python published a release: v3.1.0

08-15FLASHpypi

OpenAI's latest SDK update (3.1.0) includes hints of new `gpt-5.5` models, a `realtime` API, and mentions an upcoming Sora API shutdown.

08-14URGENTgithub

OpenAI's Codex project released a new pre-release version for its Rust implementation: 0.148.0-alpha.18.

08-14FLASHweb app

ChatGPT web app frontend deploy [025aed95]: 18 new strings (13 feature_flag, 4 endpoint, 1 model_id, 1 sentence) — verify

08-14RTNweb app

ChatGPT web app frontend deploy included 8 new strings, primarily new endpoints, indicating potential updates to image features, plugin management, and localized help/policy pages.

08-14URGENTgithub

OpenAI's codex repository published a new pre-release, rust-v0.148.0-alpha.17.

08-14RTNgithub

OpenAI's Codex repository published a new Rust pre-release, `rust-v0.148.0-alpha.16`.

08-14RTNweb app

ChatGPT web app frontend deploy [1d71c23a]: 8 new strings (8 endpoint) — verify

08-14FLASHweb app

ChatGPT web app frontend deploy [bf12bba7]: 330 new strings (155 feature_flag, 153 endpoint, 23 model_id, 15 sentence) — verify

08-14URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.14

08-14URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.15

08-14RTNweb app

ChatGPT web app frontend deploy [b7777bcf]: 8 new strings (8 endpoint) — verify

08-13FLASHweb app

ChatGPT web app frontend deploy [1f8e5b49]: 327 new strings (166 endpoint, 143 feature_flag, 19 model_id, 15 sentence) — verify

08-13RTNweb app

ChatGPT web app frontend deploy [b8e00213]: 4 new strings (4 feature_flag) — verify

08-13URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.13

08-13RTNweb app

ChatGPT web app frontend deploy [a89437bf]: 6 new strings (6 feature_flag) — verify

08-13FLASHweb app

ChatGPT web app frontend deploy [b409c184]: 330 new strings (154 feature_flag, 154 endpoint, 23 model_id, 15 sentence) — verify

08-13RTNweb app

ChatGPT web app updated with 8 new strings, including several new URLs and language-specific help and policy pages.

08-13FLASHopenapi

Two new OpenAI models, `gpt-5.5-2026-04-23` and `gpt-5.5-pro-2026-04-23`, appeared in the official API specification.

08-13URGENThacker news

HN velocity: "Accelerating GPT-5.6 Sol Ultrafast" (99 pts/hr)

08-13RTNweb app

ChatGPT web app update includes 4 new feature flags related to contentful dynamic feature pages and internal conversation rendering.

08-13FLASHweb app

ChatGPT web app frontend deploy [dcee36c5]: 10 new strings (9 feature_flag, 3 sentence, 1 model_id) — verify

08-13URGENTweb app

New strings in ChatGPT's web app suggest a GPT-4 usage cap and potential internal 'widget studio' features.

08-13FLASHweb app

ChatGPT web app frontend deploy [d8fdf332]: 328 new strings (166 endpoint, 144 feature_flag, 19 model_id, 15 sentence) — verify

08-13RTNweb app

ChatGPT web app updated with 8 new endpoint strings, potentially hinting at new features or documentation links.

08-13URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.11

08-13URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.12

08-12FLASHweb app

ChatGPT web app frontend deploy [d8fdf332]: 327 new strings (154 endpoint, 151 feature_flag, 23 model_id, 15 sentence) — verify

08-12FLASHweb app

ChatGPT web app frontend deploy [1b985495]: 328 new strings (155 endpoint, 151 feature_flag, 23 model_id, 15 sentence) — verify

08-12RTNweb app

ChatGPT web app frontend deploy [31d50d00]: 8 new strings (8 endpoint) — verify

08-12FLASHweb app

OpenAI's ChatGPT web app updated with potential references to new 'GPT-5' models, internal model access, and experimental autocomplete features.

08-12FLASHweb app

ChatGPT web app frontend deploy [b6baa400]: 12 new strings (8 feature_flag, 4 model_id) — verify

08-12RTNweb app

ChatGPT web app updated with 8 new endpoint strings, including references to images, plugins, and pricing.

08-12RTNgithub

OpenAI's Codex project published a new pre-release: rust-v0.148.0-alpha.9, suggesting ongoing development of their Rust-based tooling or SDK for code generation.

08-12RTNgithub

OpenAI Python library released v3.0.0, making HTTPX2 the default HTTP client and dropping automatic `httpx` installation, requiring migration for custom HTTPX configurations.

08-12FLASHpypi

OpenAI's new 3.0.0 SDK hints at a `gpt-4o-2024-08-06` model, and introduces 'realtime' and 'responses' beta connections, alongside JSON mode and various fine-tuning improvements.

08-12URGENTweb app

ChatGPT web app update includes new strings related to GPT-4 usage caps, internal feature flags, and a "work mode."

08-11FLASHweb app

ChatGPT web app frontend deploy [2e2dac90]: 243 new strings (121 feature_flag, 107 endpoint, 16 model_id, 15 sentence) — verify

08-11FLASHlmarena

LMArena: 1 new model on the leaderboard (solar-pro4-20260805)

08-11RTNweb app

ChatGPT web app frontend deploy [3d0c367d]: 8 new strings (8 endpoint) — verify

08-11URGENTgithub

openai/codex published a pre-release: rust-v0.148.0-alpha.8

08-11RTNgithub

openai/openai-python published a release: v2.54.0

08-11FLASHpypi

openai 2.54.0 shipped: 37 interesting new strings (36 sentence, 4 feature_flag, 2 model_id)

08-11FLASHweb app

ChatGPT web app frontend deploy [e4801150]: 6 new strings (5 feature_flag, 1 model_id, 1 sentence) — verify

08-11FLASHweb app

ChatGPT web app frontend deploy [f6781d9d]: 10 new strings (9 feature_flag, 2 sentence, 1 model_id) — verify

08-11RTNgithub

OpenAI's Codex project has released a new alpha version, rust-v0.148.0-alpha.7, indicating ongoing development in its Rust implementation.

08-11FLASHweb app

OpenAI's ChatGPT web app was updated with new strings referencing GPT-5, internal models, and experimental features like autocompletions and trending suggestions.

08-11RTNdocs

OpenAI's changelog added 'Models & Trusted Access' and 'Cyber safety' sections, replacing a more general 'Safety' entry.

08-11RTNweb app

New strings in the ChatGPT web app frontend hint at possible new features related to images, plugins, and pricing, along with links to help and a local web endpoint.

08-11FLASHweb app

New strings in ChatGPT's web app suggest a GPT-4 usage cap and new feature flags related to scheduling, automation, and internal knowledge features.

08-10FLASHweb app

OpenAI's ChatGPT web app updated with mentions of 'GPT-5', 'GPT-4.5', and 'GPT-4L', alongside several 'internal' model references and new feature flags.

08-10RTNweb app

OpenAI's ChatGPT web app deployed new frontend code with five new strings, including 'images', 'plugins', and 'pricing' endpoints.

08-10RTNgithub

OpenAI's Codex project released a new pre-release version: rust-v0.147.0-alpha.6.6.

08-10RTNweb app

ChatGPT web app update includes new strings for GPT-4 usage cap messages and several internal feature flags related to content management and analytics.

08-10FLASHweb app

OpenAI's ChatGPT web app deployed new frontend code including references to a potential 'gpt-4-5' model ID and several internal feature flags.

08-10FLASHopenapi+pages

New OpenAI models 'gpt-5.6-cyber' and 'Daybreak Blue' spotted in API spec and changelog, focused on cybersecurity tasks and defensive work, requiring special approval.

08-10RTNsitemap

OpenAI quietly added six new pages to its sitemap, including content related to cybersecurity solutions, a 'Daybreak' partner program, and premium seats for ChatGPT Business.

08-10FLASHdocs

OpenAI's changelog indicates an update to 'chat-latest' with a reference to 'GPT-5.6 Sol' for production API usage.

08-10RTNgithub

OpenAI Codex published a new pre-release for its Rust library, version 0.148.0-alpha.6.

08-16URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.56.0-nightly.20260816.g2a87e7be1

08-15URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.56.0-nightly.20260815.g2a87e7be1

08-14FLASHmodel api

New model id in Google Gemini's API: gemini-3.7-flash-video-understanding-eap

08-14URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.56.0-nightly.20260814.gc0d192452

08-13RTNhacker news

HN velocity: "Gemini 3.7 Flash" (55 pts/hr)

08-13FLASHapi_models+hn+model_catalog

OpenRouter's API, Litellm's model catalog, and an HN story all indicate the upcoming release of google/gemini-3.7-flash, and potentially a batch version.

08-13FLASHmodel api

A new model ID, gemini-3.7-flash, has appeared in Google Gemini's live API model list.

08-13URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.56.0-nightly.20260813.g1ac337739

08-12URGENTgithub

Gemini CLI pre-release v0.56.0-nightly.20260812.g5024443c7 introduces local evals reporting and fixes for model capacity exhaustion and OAuth flows.

08-11RTNgithub

Gemini CLI v0.55.1 release includes a new feature for tool registry discovery, updates to Vertex AI base URL, and various fixes.

08-11URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.56.0-preview.1

08-11URGENTgithub

google-gemini/gemini-cli published a pre-release: v0.55.0-preview.3

08-11RTNgithub

Google Gemini CLI published a new pre-release v0.56.0-nightly.20260811.geef19f25c, which includes a fix for refreshing MCP OAuth tokens.

08-10FLASHmodel api

New model IDs for Gemini 2.5 Flash with native audio preview and Gemini 3.1 Flash live preview have appeared in Google's API, indicating upcoming releases.

08-11URGENThacker news

An HN post details 11-16x faster LLM inference using Llama.cpp with Apple Silicon and macOS VMs.

08-12URGENThacker news

HN velocity: "SpaceXAI's Grok 4.6 Scores 61 on the Artificial Analysis Intelligence Index" (84 pts/hr)

08-12RTNhacker news

HN velocity: "Grok 4.6" (71 pts/hr)

08-12FLASHlmarena

LMArena has added three new models to its leaderboard, including a stealth model named 'august26-chatbot1-fmme' and two Grok-4.6 models.

08-12FLASHmodel api

A new model ID, x-ai/grok-4.6, has appeared in OpenRouter's API, suggesting a potential unannounced release.

08-11RTNhacker news

HN velocity: "Grok Bot by SpaceXAI" (66 pts/hr)

08-16FLASHmodel_catalog

Model catalog: 20 new model ids (azure_ai/FW-DeepSeek-V3.2, azure_ai/FW-DeepSeek-V4-Pro, azure_ai/FW-GLM-5…)

08-15FLASHlmarena

LMArena: 1 new model on the leaderboard (deepseek-v4-pro-high-20260813-webdev)

08-14FLASHhuggingface

Nvidia has pushed new HuggingFace repositories for 'DeepSeek-V4-Flash-nvfp4-DSpark' and 'DeepSeek-V4-Pro-nvfp4-DSpark', potentially indicating new model releases or integrations.

08-13RTNhacker news

HN velocity: "DeepSeek Harness" (49 pts/hr)

08-13FLASHhuggingface

deepseek-ai pushed a new HuggingFace repo: deepseek-ai/DeepSeek-V4-Pro-0813

08-12RTNhacker news

HN velocity: "DeepSeek V4 Pro 0813 quietly released" (52 pts/hr)

08-12FLASHapi_models+hn

New model id in OpenRouter's API: deepseek/deepseek-v4-pro-0813, qwen/qwen3.8-2.4t-a95b — corroborated across 2 sources (api_models, hn)

08-13RTNhacker news

HN velocity: "Mistral OCR 4.1" (53 pts/hr)

08-10URGENThacker news

Mistral's patent application for "Code implemented tool calls" is gaining significant traction on Hacker News, suggesting potential new developments in their function calling capabilities.

08-15FLASHmodel api

New model id in OpenRouter's API: qwen/qwen3.8-27b

08-14FLASHhuggingface

Qwen pushed a new HuggingFace repo: Qwen/Qwen3.8-27B, Qwen/Qwen3.8-27B-FP8

08-12FLASHhuggingface

Qwen pushed a new HuggingFace repo: Qwen/Qwen3.8-2.4T-A95B, Qwen/Qwen3.8-2.4T-A95B-FP8

08-13URGENTdocs

Cursor's changelog shows a new 'Builds' feature for Cloud Agents, promising faster starts and more resilient runs by pre-preparing development environments.

nvidia pushed a new HuggingFace repo: nvidia/Kimi-K3-NVFP4

Nvidia created five new HuggingFace repositories under "NVIDIA-Nemotron-Labs-Teacher-", indicating potential new model releases or datasets focused on chat, coding, reasoning, instruction following, and STEM.

A new anonymous model, 'andwise-evfd', has appeared on the LMArena leaderboard, likely indicating an unreleased, stealth model undergoing testing.

Nvidia created a new HuggingFace repository called "StormScope-NSRDB".

LMArena: 4 new models on the leaderboard (flux-3-video-20260811-i2v, hailuo-3-video-edit…) — anonymous, likely a stealth model

HN velocity: "Tim Gowers: What sort of maths are LLMs good at?" (49 pts/hr)

HN velocity: "Compression Is Prediction" (45 pts/hr)

New model id in OpenRouter's API: nvidia/nemotron-3.5-lightning

HN velocity: "Stealing Reasoning Traces from Proprietary LLM APIs" (38 pts/hr)

LMArena: 1 new model on the leaderboard (latent-sauce-79r5) — anonymous, likely a stealth model

NVIDIA released new Nemotron-3.5-Lightning-30B models on HuggingFace, including variations with different precisions and base models.

OpenRouter's API added a new model ID: nvidia/nemotron-3.5-lightning:free.

A new anonymous model, 'onyx-v1-4-33s7', has appeared on the LMArena leaderboard, potentially indicating a stealth release.

A new, anonymous model, 'lo-bah-png-uq1z', has appeared on the LMArena leaderboard, potentially indicating a stealth test of an unreleased model.

Microsoft pushed a new HuggingFace repository named microsoft/MuseVLA, potentially signaling an upcoming model or project.