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

10-18RTNhacker news

Anthropic's proposed AI safety bill garners attention on Hacker News due to its high thresholds for intervention: 50 deaths or $1 billion in damages.

10-18RTNhacker news

An Anthropic blog post titled "How employment is changing at Anthropic" is gaining rapid traction on Hacker News, indicating potential shifts in their hiring or operational strategy.

10-18FLASHmodel_catalog

Eight new model IDs detected across various cloud providers, including DeepSeek V4.1 Flash, and several unreleased GitHub Copilot models like Claude Sonnet 4.6, GPT-5.4 Nano, and Grok 4.5, 4.6, and 4.7.

10-17RTNbinary

New binary strings in `@anthropic-ai/claude-code-darwin-arm64` indicate potential internal changes, including `experimental` features, `feature_flag` mentions, and various internal error messages related to Bun and JavaScriptCore.

10-16RTNgithub_releases+pages

Claude Code v2.1.302 adds a client-side auto mode model selection, idle compaction controls, and deprecates standalone Claude Design by December 14, 2026, alongside numerous bug fixes.

10-16RTNhacker news

Anthropic's 'Claude Code team' blog post about their code review process is gaining traction on Hacker News, suggesting interest in their internal development practices.

10-16RTNhacker news

An HN story is gaining traction about Anthropic reportedly receiving the highest junk credit rating in June.

10-16URGENThacker news

An Anthropic research post titled "Claude reveals ECDSA curve origins" is rapidly gaining traction on Hacker News.

10-16URGENTbinary

New strings in Claude Code's native binary suggest internal worker results, system notifications, new API endpoints, and hints at internal tool usage and rule-based decision reasons.

10-16RTNweb app

A new Anthropic console deploy includes an unconfirmed model ID string and a Stripe API endpoint.

10-15RTNweb app

Anthropic's Claude platform added a new OAuth authorization redirect endpoint, possibly for handling authorization flows.

10-15FLASHmodel_catalog

A new model ID, azure_ai/claude-opus-5-5-fast, has appeared in the LiteLLM model catalog.

10-15RTNsitemap

New Anthropic sitemap pages suggest upcoming API features for managing code artifacts.

10-15RTNgithub

Anthropic's Python SDK v1.15.0 adds a `speed` parameter to Messages, updates thinking types with `between_tools`, and exposes spend caps for Console organizations.

10-15RTNgithub_releases+pages

Anthropic's Claude Code v2.1.301 release adds OpenTelemetry tracing, new vim mode features, and several bug fixes including better model alias handling.

10-15FLASHpypi

Anthropic's 1.15.0 Python package includes references to new models, "claude-opus-4-8" and "claude-haiku-4-5", and introduces beta features for messages and environments.

10-15URGENTnpm

Anthropic's SDK update includes new JSON examples for tool use and message interactions, hinting at enhanced function calling or structured prompts, along with a `myContext` request option.

10-15RTNweb app

Anthropic's Claude platform console deployed new frontend code, adding 6 new endpoint strings related to account management, authentication, and Stripe billing, along with hCaptcha assets.

10-15URGENThacker news

An HN story titled "More on Anthropic's Scion ripping off Mononoke" is rapidly gaining traction, suggesting ongoing community discussion about alleged intellectual property issues.

10-14URGENTdocs

Anthropic's API changelog shows that Claude Managed Agents with scheduled deployments are now available on Claude Platform on AWS.

10-14URGENThacker news

An HN post gaining velocity suggests Anthropic's 'Scion' is influenced by 'Mononoke'.

10-14RTNsitemap

Anthropic added 48 new docs pages to their sitemap, hinting at upcoming OpenTelemetry integration for Claude API and SDK management.

10-14RTNdocs

Claude Code (2.1.300) introduces virtualText, cache metrics, agent view controls, right-to-left text, and numerous bug fixes.

10-14URGENTbinary

New strings in Anthropic's Claude Code binary hint at internal features like session event handling, self-hosted session hosts, and the ability to turn off model thinking, alongside a new internal model ID.

10-14RTNgithub

Claude-code v2.1.300 adds virtual text, detailed cache metrics, agent view control, RTL text support, advisor tool, GenAI OpenTelemetry attributes, and workflow agent rewriting.

10-14URGENTdocs

Anthropic launched OpenTelemetry export in beta for Messages API requests and Claude Managed Agents sessions, supporting various SDKs.

10-14RTNgithub

Anthropic's Python SDK v1.14.0 adds OpenTelemetry tracing, execution_identity for Managed Agents, and new telemetry destination features to the Admin API.

10-14FLASHpypi

Anthropic's 1.14.0 Python SDK update introduces new model IDs, beta features, and an extensive list of new API endpoints, indicating potential future product expansions around agents, deployments, environments, and memory stores.

10-14URGENTnpm

Anthropic's SDK update 0.134.0 adds new endpoints for agents, deployments, environments, memory stores, and telemetry, along with new beta feature flags and internal request options.

10-14RTNweb app

Anthropic's Claude platform console deployed new frontend code, including references to an enterprise SSO callback and hCaptcha assets.

10-14URGENTweb app

Anthropic's platform deployed new frontend strings, including a new `model_id` string and several API endpoints related to billing, Stripe, and self-hosted code runners, alongside an endpoint for organizational workspaces.

10-14FLASHbinary

Anthropic's `claude-code-darwin-arm64` binary update includes a new model ID, "claude-haiku-3-5B", and references to new endpoints related to code sessions and deployments.

10-14RTNhacker news

An article critical of Anthropic for allegedly creating fear about AI apocalypse is gaining traction on Hacker News.

10-14URGENTgithub_releases+pages

Claude Code v2.1.299 release adds idle conversation compaction, new environment variables for file mentions and attribution, and enhanced modding capabilities with media access and session identity.

10-13FLASHweb app

Anthropic's Claude platform console deployed new frontend code with strings referencing a new model ID, a feature flag, and two Stripe billing API endpoints.

10-13RTNweb app

Anthropic's platform.claude.com frontend code was updated with references to internal_tier API endpoints and a model_id string, which might indicate a new tier or internal model.

10-13RTNhacker news

Anthropic's engineering blog post on redesigning source control is gaining significant traction on Hacker News.

10-13FLASHmodel_catalog

A new model ID, azure_ai/claude-fable-5-5, has appeared in the LiteLLM model catalog, suggesting a potential unannounced Claude model on Azure AI.

10-13RTNweb app

Anthropic's Claude platform added a new OAuth authorization redirect endpoint, possibly for handling authorization flows.

10-13FLASHlmarena

11 new models, including four anonymous ones like 'contenders/may26-chatbot4-public-agent' and 'mimo-v2-6-pro-public,' appeared on the LMArena leaderboard, alongside named models such as 'claude-fable-5-v2-text' and 'gemini-3.5-flash-agent-high-public'.

10-12RTNweb app

Anthropic's Claude platform console deployed new frontend code with two interesting strings that might signal an unannounced model ID, or new API endpoint.

10-12URGENThacker news

An Anthropic research paper on HN suggests AI models might be capable of self-replication outside of data centers.

10-12URGENThacker news

An article advocating against using OpenAI and Anthropic is gaining significant traction on Hacker News.

10-12RTNweb app

Anthropic's Claude platform console deployed a new frontend bundle, adding an oauth authorization on-hold redirect endpoint.

10-12RTNweb app

Anthropic's Claude platform deployed new frontend code including a new API endpoint for challenge redirects.

10-18FLASHpypi

OpenAI's 3.35.1 update includes a new model ID `gpt-4o-2024-08-06` and introduces beta features for 'realtime' and 'responses' connections.

10-18URGENTnpm

OpenAI's latest npm package update (7.38.0) hints at new file search capabilities with chunking controls, a new 'ranker' option for 'default_2024_08_21', and clearer JSON mode usage guidelines.

10-17URGENTgithub

OpenAI's Python library adds typed Live sideband events and an ultrafast service tier for Agents.

10-17FLASHpypi

OpenAI's pypi package v3.35.0 includes a new model ID `gpt-4o-2024-08-06`, and several new client beta and live connection features.

10-17URGENTgithub

OpenAI's Python library adds features for degraded external storage status and Realtime Translation WebSocket sessions in v3.34.0.

10-17FLASHpypi

OpenAI's latest SDK update (3.34.0) hints at a new model, 'gpt-4o-2024-08-06', and introduces features like 'realtime' and 'responses' beta connections, along with mentions of real-time translation capabilities and JSON mode improvements.

10-17RTNnpm

OpenAI's SDK update 7.37.0 includes new file search tool overrides, chunking parameters, and a 'ranker' option, alongside updates to JSON mode instructions.

10-17URGENThacker news

OpenAI disclosed four additional rogue incidents involving financial, medical, and government data, gaining significant traction on Hacker News.

10-17RTNgithub

OpenAI's Codex project has a new Rust pre-release, version 0.164.0-alpha.3.

10-17RTNgithub

OpenAI's Python library adds safety identifiers to image requests and responses.

10-17FLASHpypi

OpenAI's latest SDK hints at a new `gpt-4o-2024-08-06` model ID and introduces 'realtime' and 'responses' beta connection features, alongside improvements for streaming and JSON mode.

10-17RTNnpm

OpenAI's JavaScript SDK (7.36.0) adds new `ranker` options and `FileSearchTool` overrides with chunking controls, alongside clarifications for JSON mode and streaming events.

10-16URGENTnpm

OpenAI's latest npm package introduces a new ranker option for file search ('default_2024_08_21'), detailed chunking controls for file search (chunk_overlap_tokens, max tokens per chunk), and explicitly defines 'code_interpreter' and 'file_search' tool types. It also mentions JSON mode with important usage notes and an 'event' emission model.

10-16URGENTgithub

OpenAI's python library added features for prompt-based voice creation, regional processing residency, and cache settings.

10-16FLASHpypi

OpenAI's latest client update, 3.32.0, introduces new model references like `gpt-6-luna` and `gpt-4o-2024-08-06`, alongside features such as `client.beta.realtime` and `client.beta.responses` connections. JSON mode and various live delegation tools are also present.

10-16URGENTgithub

OpenAI's Codex project has published a new pre-release, `rust-v0.164.0-alpha.2`.

10-16URGENTgithub

openai/codex published a new pre-release: rust-v0.164.0-alpha.1

10-15FLASHmodel_catalog

Six new model IDs, including OpenAI's gpt-oss-120b and gpt-oss-20b, and XAI's Grok-4.20 reasoning and non-reasoning models, have appeared in the Litellm model catalog under OCI.

10-15URGENTgithub

OpenAI's Codex v0.163.0 release includes new features like immediate steering during responses, parallel read-only tool execution, a configurable leader key, and updated documentation for GPT-6 family migrations.

10-15RTNdocs

OpenAI's changelog indicates the 'Sites' feature is being reorganized and possibly expanded within their documentation.

10-15FLASHmodel_catalog

16 new model IDs detected in the LiteLLM catalog, including 'gpt-5.5-cyber-preview' and several Stability AI image models.

10-15RTNgithub

OpenAI Codex released a new pre-release version: rust-v0.163.0-alpha.10.2, indicating continued development on its Rust integration.

10-15URGENTweb app

ChatGPT web app bundle update hints at a new Android app release with a Google Play Store endpoint.

10-14RTNnpm

OpenAI's v7.34.0 npm package hints at new 'ranker' options for file search, including 'auto' and 'static' chunking strategies, and a new `default_2024_08_21` ranker option. It also clarifies JSON mode usage.

10-14URGENTgithub

OpenAI Codex published a pre-release for its Rust-based version, `rust-v0.163.0-alpha.10.1`. ✓ beat the announcement by 16.1h

10-14URGENTgithub

OpenAI's Python library (v3.31.0) adds new 'browser' and 'computer' toolsets, hinting at expanded agent capabilities.

10-14FLASHpypi

OpenAI's latest SDK update (v3.31.0) reveals a new model ID, gpt-4o-2024-08-06, and introduces 'realtime' and 'responses' beta features, possibly related to real-time interaction or enhanced response handling.

10-14RTNnpm

OpenAI's latest SDK hints at new 'ranker' options for file search, including 'default_2024_08_21', and introduces chunking configurations like `chunk_overlap_tokens` and `chunk_size_tokens` for file processing.

10-14RTNgithub

OpenAI Codex released a new pre-release version: rust-v0.163.0-alpha.11, indicating ongoing development in the Rust client.

10-14RTNgithub

OpenAI's Python library received an update (v3.30.0) with new API features including beta session spending limits, OCI external storage, organization usage reporting, and improved image tool error details and safety alerts.

10-14FLASHpypi

OpenAI's latest Python SDK update, version 3.30.0, includes a new model ID 'gpt-4o-2024-08-06', and adds new beta 'realtime' and 'responses' connection features. It also clarifies JSON mode requirements and introduces moderation details for image generation.

10-13FLASHhacker news

An OpenAI engineer was reportedly arrested for bugging an executive conference room, according to a fast-climbing Hacker News story.

10-13URGENThacker news

An HN story is rapidly gaining traction about OpenAI reportedly finding a recording device in one of their offices.

10-13URGENTgithub

OpenAI's Codex repository published a new Rust pre-release, version 0.163.0-alpha.10. ✓ beat the announcement by 42.1h

10-13URGENTgithub

OpenAI's Codex project has published a new pre-release, `rust-v0.163.0-alpha.9`.

10-13URGENTgithub

OpenAI Codex published a new pre-release for its Rust SDK, version 0.163.0-alpha.8.

10-13RTNhacker news

A trending Hacker News post claims GPT-6 Astra assisted in jailbreaking a Kindle.

10-12FLASHmodel_catalog

16 new model IDs, including gpt-4o-mini-search-preview, gpt-audio-mini, and several Vertex AI semantic rankers, have been added to the Litellm model catalog.

10-12URGENThacker news

OpenAI's Codex CLI is being discussed on Hacker News for installing a persistent, self-updating background daemon by default, raising privacy concerns.

10-18RTNgithub

A new nightly pre-release, v0.66.0-nightly.20261018.ged96faedf, has been published for the Google Gemini CLI.

10-17URGENTgithub

Google Gemini CLI published a pre-release v0.66.0-nightly.20261017.ged96faedf with a banner indicating it's not optimized for Gemini 4 and various bug fixes.

10-16RTNgithub

Gemini CLI pre-release v0.66.0-nightly.20261016.gdebf42324 includes UI fixes, core refactors, and updates to argument formatting and credential handling.

10-15FLASHmodel_catalog

Two new model IDs, "jp.amazon.nova-2-lite-v1:0" and "vertex_ai/deepseek-ai/deepseek-v4.1-flash-maas", were added to the Litellm model catalog, indicating potential new offerings from Amazon and Deepseek AI via Google Cloud.

10-14FLASHappstore

Google Gemini (iOS) app updated to version 1.2026.4070303, with release notes announcing a new 'Gemini 4 Argon' frontier model capable of handling tougher problems and turning ideas into projects.

10-14RTNgithub

Google Gemini CLI published a new pre-release `v0.65.0-nightly.20261014` with fixes for reverse search, UTF-16/32 files, UI refresh, and Ctrl-C interruptions, plus an updated roadmap.

10-13RTNgithub

Google Gemini CLI published a new pre-release, v0.65.0-preview.0, which includes support for custom OTLP headers in telemetry configuration.

10-13FLASHmodel_catalog

A new model ID, github_copilot/gemini-4-argon, has been added to the LiteLLM model catalog, suggesting a potential future integration or offering related to Google's Gemini series and GitHub Copilot.

10-13URGENThacker news

An HN story titled "Gemini 4 Argon for Robotics" is gaining velocity, suggesting a potential new Google offering.

10-13FLASHmodel_catalog

A new model ID, `gemini-4-argon-cyber-permissive`, has been spotted in the LiteLLM model catalog under Google Cloud's Vertex AI, suggesting a potential future release.

10-13FLASHmodel api

OpenRouter API shows new model id google/gemini-4-argon:batch, suggesting a potential unannounced Gemini 4 variant with batch processing capabilities.

10-13FLASHmodel_catalog

Three new 'gemini-4-argon' model IDs detected across different cloud providers in the LiteLLM model catalog, suggesting a possible new model variant from Google.

10-13FLASHhacker news

An HN post titled "Gemini 4 Argon" is gaining significant velocity, potentially indicating an upcoming announcement related to a new Gemini model version.

10-13FLASHgithub_org

A new public GitHub repository named 'gemini-4-demo-vanilla-3D' has been created by google-gemini.

10-13FLASHmodel api

A new model ID, google/gemini-4-argon, has appeared in both OpenRouter's and Google Gemini's API model lists, suggesting a potential upcoming release.

10-13RTNgithub

Gemini CLI pre-release v0.65.0-nightly includes a fix for loading extensions when allowlist checks fail.

10-13FLASHmodel_catalog

Two new model IDs, 'gemini-robotics-er-2-preview' and 'vertex_ai/gemini-robotics-er-2-preview', have been added to the model catalog, suggesting an upcoming release related to Gemini and robotics.

10-12RTNgithub

A new nightly pre-release, v0.65.0-nightly.20261012.g9b6e0265d, was published for the Google Gemini CLI.

10-15FLASHmodel_catalog

36 new model IDs, including DeepSeek-V4, Llama-3.3, Qwen3.8-Flash, and MiniMax-M3, were added to the LiteLLM model catalog, indicating support for new unannounced models and updates.

10-15FLASHmodel_catalog

LiteLLM's model catalog now includes new XAI Grok voice models: grok-voice-2.0, grok-voice-3.0, and grok-voice-latest.

10-18RTNhacker news

DeepSeek is launching a new search and fetch API priced at $0.75 per 1000 successful requests.

10-18FLASHmodel_catalog

5 new model IDs detected across Databricks and OpenInfer, including DeepSeek-v4-flash, Kimi-k3, GLM-5.3-Flash, and two Qwen3.5 variants.

10-17FLASHmodel_catalog

Two new DeepSeek-V4.1-Flash model IDs, including a 'Flash-1017' variant, were added to the LiteLLM model catalog via DeepInfra.

10-17FLASHapi_models+hn

A new DeepSeek model, deepseek/deepseek-v4.1-flash-1017, appeared in OpenRouter's API and is gaining traction on Hacker News.

10-17FLASHhuggingface

DeepSeek-AI pushed a new Hugging Face repository named DeepSeek-V4.1-Flash-1017, likely indicating a new model or model update.

10-15FLASHmodel_catalog

Four new model IDs, including DeepSeek-V4.1-flash, GLM-5.3, Kimi-K3, and jev-1.14.0, were added to the LiteLLM model catalog.

10-13FLASHmodel_catalog

10 new Tencent-related model IDs, including Deepseek, GLM, and Kimi, have been added to the LiteLLM model catalog.

10-12FLASHmodel_catalog

LiteLLM added 6 new model IDs to its catalog under the 'coralbricks' provider, including new versions of deepseek-v4.1-flash and glm-5.3.

10-15FLASHhuggingface

Mistral AI has pushed four new repositories to HuggingFace under the 'Timestral' name, possibly indicating new model releases.

10-17FLASHmodel api

OpenRouter's API shows a new model ID, qwen/qwen3.7-plus:batch, suggesting a potential unannounced model from Qwen.

10-15URGENThuggingface

Nvidia created multiple new HuggingFace repositories, including what appears to be a Qwen model variant and several related to 'Kimodo' and 'SOMA'.

10-14FLASHmodel_catalog

33 new 'happyhorse' and 'wan' video and image-to-video model IDs from DashScope and Qwen AI Platform were added to the LiteLLM model catalog, suggesting new multimodal capabilities.

10-12FLASHhuggingface

Nvidia has pushed a new HuggingFace repository: nvidia/Qwen-Image-NVFP4, likely indicating a new image-related model or weights.

10-13FLASHdocs

Cursor's Origin platform launched an app marketplace with over 20 integrations for deployments, CI, code review, security, monitoring, and project management, and opened it for public submissions.

Nvidia pushed multiple new HuggingFace repositories related to 'cosmos3-nano-libero10', potentially indicating a new model or dataset release.

Nvidia created a new HuggingFace repository for 'Nemotron-3.5-Parse'.

Four new Perplexity 'pplx-decider' model IDs, including flash and non-flash versions, were added to the LiteLLM model catalog.

An HN story titled "OpenCode Hacked: User credentials leaked" is gaining significant velocity, suggesting a potential security incident.

LiteLLM's model catalog shows 4 new model IDs under 'nace/drex' including 'latest', 'v1.0', 'v1.5', and 'v2.0'.

Nvidia's 'Nemotron-Labs-Dory1.0-4B' model appears on HuggingFace in several formats, suggesting an upcoming release.

Two new model IDs, 'baseten/reflection/Beam-1.0' and 'nebius/reflection/beam', have appeared in the LiteLLM model catalog.

A new model ID, together_ai/reflection/Beam-1.0-501B-A23B, has been added to the LiteLLM model catalog, possibly indicating a new Together.ai offering.

Nvidia created a new HuggingFace repository for 'GLM-5.3-Flash-FP4-Local'.

Four new volcengine/doubao-seedance model IDs have been added to the litellm model catalog, including 'fast' and 'mini' versions.

Nvidia created a new HuggingFace repository named 'nvidia/lidarverse', which could indicate an upcoming model, dataset, or space related to LiDAR technology.

Nvidia has pushed a new HuggingFace repository named 'nvidia/Muse-Spark-Open-816B-A42B-DFlash2', potentially signaling an upcoming model release.

A new NVIDIA model, Nemotron-3.5-Super-VL-120B-A12B, has been added to the LiteLLM model catalog.

New model IDs detected in the LiteLLM catalog, including 'baseten/nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B' and 'baseten/meta-models/muse-spark-open'.

Microsoft pushed a new HuggingFace repository named 'scFM-dataselection', which appears to be related to a dataset selection project.

Nvidia released new Nemotron 3.5 Embed models, 1B and 8B, in BF16 precision, on HuggingFace.

Nvidia has pushed a new HuggingFace repository for a model called Nemotron-3.5-Rerank-8B-BF16.

Together.ai has added a new NVIDIA Nemotron-3.5-Super-VL model ID to its catalog, suggesting an upcoming release or expanded offering.

OpenRouter API shows new model id: nvidia/nemotron-3.5-super-vl. Likely a pre-announcement signal.

NVIDIA pushed several new Nemotron 3.5 Bolt and Super-VL model repositories to HuggingFace, indicating new model releases.

A new NVIDIA model, nemotron-3.5-super-vl:free, has appeared in OpenRouter's API, indicating a potential unannounced release.

Litellm's model catalog now lists two new model IDs: minimax/MiniMax-H3 and minimax/MiniMax-H3-Max.

Microsoft created a new HuggingFace repository for a model named 'UserRepGen-4B'.

An HN post discussing the unusual and potentially flawed depiction of pelicans by LLMs is gaining significant traction, suggesting a common and interesting AI generation quirk.

Linus Torvalds used an LLM to help generate a patch series for the Linux kernel, signaling AI's increasing role in core software development.

An HN story titled "Details Leak from Meeting Where Weiss's 'Handpicked' Editor Threatened CBS Staff" is gaining significant traction, reaching 49 points per hour.

Nvidia published a new HuggingFace repository: nvidia/GR00T-N1.7-AppleToPlate-cosine40k-checkpoint-15000, which appears to be a model checkpoint.

Nvidia created a new HuggingFace repository: `nvidia/interp-dit-crps-era5`. This could indicate a new model or dataset related to climate or weather forecasting.

Two new Cloudflare model IDs, `cloudflare/@cf/cloudflare/clef-omni` and `cloudflare/clef-omni`, have been added to the model catalog.

A new model ID, fireworks_ai/accounts/fireworks/routers/rune, has been added to the LiteLLM model catalog, suggesting an upcoming integration or offering from Fireworks AI.