anthropics/claude-code v2.1.303
Anthropic releases Claude Code version 2.1.303, which includes bug fixes and reliability improvements. ·
New release v2.1.303 of anthropics/claude-code on GitHub.
Top now
6 storiesWhat is climbing fastest across every source
Anthropic6.1xLLM9.9xDeepSeek7.2xOpenAI4.9xV4.16.1xGoogle5.4xDeepSeek-V4.1-Flash-10174.1xFlash-10174.1xMeta3.0x
[Industry Event] Advancing Data Center Construction East 2026
From Data Center Knowledge.
Show HN: WhatsApp MCP running free on Cloudflare
A read-only MCP server lets AI agents search and read personal WhatsApp messages while running on Cloudflare Workers. ·
Hybrid Reasoning Agents: LLM agents that think in latent space, act through text
Researchers introduce HybridAgents, which dynamically switch between latent reasoning and text-based interaction to bridge the gap between internal efficiency and external tool access. ·
reflection/Beam-1.0-501B-A23B
A text generation model from reflection with 148 likes and 1,591 downloads on Hugging Face.
alchaincyf/huashu-motion
The repository alchaincyf/huashu-motion lets coding agents generate commercial product launch videos and allows users to edit specific frames directly. ·
anthropics/claude-code v2.1.303
Anthropic releases Claude Code version 2.1.303, which includes bug fixes and reliability improvements. ·
New release v2.1.303 of anthropics/claude-code on GitHub.
Hybrid Reasoning Agents: LLM agents that think in latent space, act through text
Researchers introduce HybridAgents, which dynamically switch between latent reasoning and text-based interaction to bridge the gap between internal efficiency and external tool access. ·
Show HN: WhatsApp MCP running free on Cloudflare
A read-only MCP server lets AI agents search and read personal WhatsApp messages while running on Cloudflare Workers. ·
alchaincyf/huashu-motion
The repository alchaincyf/huashu-motion lets coding agents generate commercial product launch videos and allows users to edit specific frames directly. ·
vllm-project/vllm v0.32.0
vLLM releases version 0.32.0, featuring 814 commits from 363 contributors, including new structured decision endpoints and low-precision KV cache support. ·
deepseek-ai/DeepSeek-V4.1-Flash-1017
A image text to text model from deepseek-ai with 1,330 likes and 10.2k downloads on Hugging Face.
deepseek-ai/DeepSeek-V4.1-Flash-1017
A image text to text model from deepseek-ai with 1,330 likes and 10.2k downloads on Hugging Face.
anthropics/claude-code v2.1.303
Anthropic releases Claude Code version 2.1.303, which includes bug fixes and reliability improvements. ·
Show HN: WhatsApp MCP running free on Cloudflare
A read-only MCP server lets AI agents search and read personal WhatsApp messages while running on Cloudflare Workers. ·
Anthropic's proposed AI safety bill only kicks in if 50 deaths or $1B in damages
115 points and 129 comments on Hacker News, linking tomshardware.com.
alchaincyf/huashu-motion
The repository alchaincyf/huashu-motion lets coding agents generate commercial product launch videos and allows users to edit specific frames directly. ·
vllm-project/vllm v0.32.0
vLLM releases version 0.32.0, featuring 814 commits from 363 contributors, including new structured decision endpoints and low-precision KV cache support. ·
Trending topics today: heat
Table
| Anthropic | 1.73 |
| LLM | 1.33 |
| DeepSeek | 0.72 |
| OpenAI | 0.72 |
| V4.1 | 0.67 |
| 0.54 |
Heat = summed hotness of today's stories naming the topic.
Models
5 storiesReleases, papers, trending weights and new listings
Top 6 trending on Hugging Face: likes
Table
| google/embeddinggemma-2 ▼2 | 2661 |
| jialinyyzz/humanizer ▼1 | 1946 |
| deepseek-ai/DeepSeek-V4.1-Flash-1017 ▲1 | 1330 |
| openbmb/MiniCPM-V-4.7-35B-A3B ▲2 | 1007 |
| black-forest-labs/flux-3-image-klein-comfy | 806 |
| fishaudio/drama-3 | 690 |
PirateBench: score out of 100
Table
| gpt-6.1-sol | 98.9 |
| opus-5-5 | 96.46 |
| gpt-6-sol | 91.6 |
| gpt-6-luna | 65.25 |
| haiku-4-5 | 4.28 |
Agents build our own projects from a brief; every run, failures included: https://cowerx.dev/daily/piratebench/
DeepSeek search and fetch – $0.75 per 1000 successful requests
DeepSeek charges $0.75 per 1000 successful requests for its Search and Fetch APIs, which provide agents and RAG applications access to live web content. ·
529 points and 252 comments on Hacker News, linking api-docs.deepseek.com.
deepseek-ai/DeepSeek-V4.1-Flash-1017
A image text to text model from deepseek-ai with 1,330 likes and 10.2k downloads on Hugging Face.
Comfy-Org/MOSS-Image
A model model from Comfy-Org with 190 likes and 354.1k downloads on Hugging Face.
nvidia/NVIDIA-Nemotron-3.5-Super-VL-120B-A12B-NVFP4
A image text to text model from nvidia with 155 likes and 7,599 downloads on Hugging Face.
DeepSeek: DeepSeek v4.1 Flash 1017
DeepSeek serves the October 17, 2026 snapshot of its V4.1 Flash model under the deepseek-flash API name. ·
deepseek-ai/DeepSeek-V4.1-Flash-1017
A image text to text model from deepseek-ai with 1,330 likes and 10.2k downloads on Hugging Face.
DeepSeek: DeepSeek v4.1 Flash 1017
DeepSeek serves the October 17, 2026 snapshot of its V4.1 Flash model under the deepseek-flash API name. ·
DeepSeek search and fetch – $0.75 per 1000 successful requests
DeepSeek charges $0.75 per 1000 successful requests for its Search and Fetch APIs, which provide agents and RAG applications access to live web content. ·
Comfy-Org/MOSS-Image
A model model from Comfy-Org with 190 likes and 354.1k downloads on Hugging Face.
OpenMOSS-Team/MOSS-Image-Base
A text to image model from OpenMOSS-Team with 289 likes and 1,791 downloads on Hugging Face.
X voices
5 storiesX posts other sources are linking to, then named voices on Bluesky and Mastodon
@beeechimblind
Israel makes propaganda video featuring AI generated IAMDDB without her consent
27 points and 2 comments on Hacker News, linking an X post.
@anemll
Qwen3.8-27B on the iPhone 18 Neural Engine at 15–70 tok/s (1.5 bpw)
30 points and 12 comments on Hacker News, linking an X post.
@onloglogn
Apple releases RSI-Inference: evaluating agents for self-improving LLM inference
2 points and 0 comments on Hacker News, linking an X post.
@deepseek_ai
DeepSeek-v4.1-Flash-1017 is here
9 points and 0 comments on Hacker News, linking an X post.
@worldnetwork
Muse, dots, Instinct, and the question every agent will be asked
1 points and 0 comments on Hacker News, linking an X post.
@anemll
Qwen3.8-27B on the iPhone 18 Neural Engine at 15–70 tok/s (1.5 bpw)
30 points and 12 comments on Hacker News, linking an X post.
@beeechimblind
Israel makes propaganda video featuring AI generated IAMDDB without her consent
27 points and 2 comments on Hacker News, linking an X post.
@deepseek_ai
DeepSeek-v4.1-Flash-1017 is here
9 points and 0 comments on Hacker News, linking an X post.
@onloglogn
Apple releases RSI-Inference: evaluating agents for self-improving LLM inference
2 points and 0 comments on Hacker News, linking an X post.
GitHub
5 storiesNew repos, releases and changelog
New AI repos this week: stars
Table
| feder-cr/git-ai | 2293 |
| deeplethe/solid-glass | 2015 |
| zhuyansen/new-term-radar | 1108 |
| outcrawlai/scout | 945 |
| Leonxlnx/earshot | 793 |
| LiCinCan/blender-ai-mcp-addon | 785 |
vllm-project/vllm v0.32.0
vLLM releases version 0.32.0, featuring 814 commits from 363 contributors, including new structured decision endpoints and low-precision KV cache support. ·
New release v0.32.0 of vllm-project/vllm on GitHub.
Codex Desktop continuously generates Crashpad pending dumps
Codex Desktop continuously generates Crashpad pending dumps, adding paired files to a directory that grew to 4.9 gigabytes on one user's machine. ·
aicodeking/leanrea
A developer publishes an eight-kilobyte skill that replaces the Reverse Engineer Anything tool using plain command-line utilities, claiming it runs 38% faster and costs 35% less. ·
openai/openai-python v3.35.1
OpenAI releases version 3.35.1 of its Python library, which fixes a bug that preserved beta Realtime byte reads on websockets. ·
openai/openai-python v3.35.0
OpenAI releases version 3.35.0 of its Python library, adding typed Live sideband events and an ultrafast service tier for Agents. ·
vllm-project/vllm v0.32.0
vLLM releases version 0.32.0, featuring 814 commits from 363 contributors, including new structured decision endpoints and low-precision KV cache support. ·
New release v0.32.0 of vllm-project/vllm on GitHub.
aicodeking/leanrea
A developer publishes an eight-kilobyte skill that replaces the Reverse Engineer Anything tool using plain command-line utilities, claiming it runs 38% faster and costs 35% less. ·
Codex Desktop continuously generates Crashpad pending dumps
Codex Desktop continuously generates Crashpad pending dumps, adding paired files to a directory that grew to 4.9 gigabytes on one user's machine. ·
anthropics/claude-code v2.1.302
Anthropic releases Claude Code version 2.1.302, which adds configuration options for auto mode and idle compaction, and fixes several bugs related to memory and model switching. ·
ollama/ollama v0.40.6
New release v0.40.6 of ollama/ollama on GitHub.
Infrastructure
5 storiesData centers, power, chips and what tokens cost
Data-center and energy stories per day
Table
| Tue 13 | 29 |
| Wed 14 | 34 |
| Thu 15 | 41 |
| Fri 16 | 34 |
| Sat 17 | 5 |
| Sun 18 | 2 |
| Mon 19 | 0 |
Stories from DCD, Data Center Knowledge, EIA and Hacker News, by publication day.
Median output price of models added in the last 30 days, $ per million tokens
One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.
Output price, $ per million tokens (newest models)
Table
| deepseek-v4.1-flash-1017 | 0.6 |
| glimpse | 3 |
| beam-1.0-501b-a23b | 3 |
| moondream3-preview | 2.5 |
| moondream3.1-9b-a2b | 2.5 |
| median, all paid models | 2 |
443 paid models listed; median of the 49 added in 30 days: $3.20. Rankings data by OpenRouter, CC BY 4.0.
The end of independent infrastructure roadmaps: how AI is forcing data center co-design
AI workloads force data centers to co-design infrastructure, preventing idle accelerators from wasting energy and reducing returns on expensive investments. ·
From DCD.
Why data sovereignty is about control, not geography
Data sovereignty centers on control rather than physical location, and the distinction between residency and sovereignty causes trouble for most organizations. ·
Plans for data center in Moncalieri, Italy, put on hold by town mayor
The mayor of Moncalieri, Italy, pauses a proposed data center project that was described as the country's largest. ·
Why harnesses matter for energy (and token) efficiency
A month of agent traffic indicates that the harness, rather than the model alone, determines the energy required for a coding session. ·
46 points and 23 comments on Hacker News, linking neuralwatt.com.
Market selloff, rising rates test AI data centre deal boom
49 points and 40 comments on Hacker News, linking reuters.com.
The end of independent infrastructure roadmaps: how AI is forcing data center co-design
AI workloads force data centers to co-design infrastructure, preventing idle accelerators from wasting energy and reducing returns on expensive investments. ·
A story of WiFi secrets, UART lockdowns and a damaged flash memory chip
Researchers reverse engineer the Eufy Homebase2, revealing its WiFi secrets, UART lockdowns, and a damaged flash memory chip. ·
