Archived front page, Fri 09 Oct 2026, 01:00 UTC. All of 2026-10-09 · Archive · Live page

The Daily Markdown

The front page of the internet for AI users · Fri 09 Oct 2026, 01:00 UTC · Get it by email

Top now

6 stories

What is climbing fastest across every source

OpenAI5.6xClaude Code7.6xBenchmark7.1xClaude Haiku 5.57.0xScaling8.1xClaude4.9xOpus 5.55.9xAnthropic6.6xFinland5.7xChatGPT5.3x

OpenAI, the Partition Principle, and Mathematics

OpenAI announces that the Partition Principle does not hold, a claim that prompted multiple inquiries to the author regarding the mathematical implications of the statement. ·

Hacker News2h ago▲5 to 125 across 5 pollsdiscusskaragila.org

langchain==1.4.4

LangChain releases version 1.4.4, which adds a retry mechanism for the summarization middleware when context overflow occurs. ·

GitHub Releases3h ago▲github.com

ThariqS/ai-newtab

A new repository written in TypeScript with 171 stars in its first week.

GitHub7h ago▲135 to 171 across 2 pollsgithub.com

Permits, Not Power, Halt Google’s Finland Data Centers

Environmental regulations block construction at two sites in Google's €13 billion Finland data center expansion, despite the company securing access to the northern grid. ·

Data Center Knowledge7h ago▲datacenterknowledge.com

OpenAI, the Partition Principle, and Mathematics

OpenAI announces that the Partition Principle does not hold, a claim that prompted multiple inquiries to the author regarding the mathematical implications of the statement. ·

67 points and 58 comments on Hacker News, linking karagila.org.

Hacker News2h ago▲5 to 125 across 5 pollsdiscusskaragila.org

ThariqS/ai-newtab

A new repository written in TypeScript with 171 stars in its first week.

GitHub7h ago▲135 to 171 across 2 pollsgithub.com

openai/codex 0.162.0

OpenAI releases Codex version 0.162.0, which adds Git worktree management tools and enables users to pin tasks in the agent Command Center. ·

GitHub Releases6h ago▲0 to 2 across 3 pollsgithub.com

autotrust/GLM5.3-Flash-E224-DGX-Spark

A image text to text model from autotrust with 396 likes and 5,422 downloads on Hugging Face.

Hugging Face15h ago▲132 to 396 across 31 pollshuggingface.co

OpenAI, the Partition Principle, and Mathematics

OpenAI announces that the Partition Principle does not hold, a claim that prompted multiple inquiries to the author regarding the mathematical implications of the statement. ·

67 points and 58 comments on Hacker News, linking karagila.org.

Hacker News2h ago▲5 to 125 across 5 pollsdiscusskaragila.org

Claude Haiku 5.5

1,032 points and 482 comments on Hacker News, linking anthropic.com.

Hacker News+ Simon Willison31h ago▲72 to 1514 across 132 pollsdiscussanthropic.com

ThariqS/ai-newtab

A new repository written in TypeScript with 171 stars in its first week.

GitHub7h ago▲135 to 171 across 2 pollsgithub.com

autotrust/GLM5.3-Flash-E224-DGX-Spark

A image text to text model from autotrust with 396 likes and 5,422 downloads on Hugging Face.

Hugging Face15h ago▲132 to 396 across 31 pollshuggingface.co

openai/codex 0.162.0

OpenAI releases Codex version 0.162.0, which adds Git worktree management tools and enables users to pin tasks in the agent Command Center. ·

GitHub Releases6h ago▲0 to 2 across 3 pollsgithub.com

Trending topics today: heat

OpenAI: 2.82OpenAI2.82Claude Code: 1.91Claude Code1.91Benchmark: 1.23Benchmark1.23Claude Haiku 5.5: 1.02Claude Haiku 5.51.02Scaling: 0.93Scaling0.93Claude: 0.82Claude0.82
Table
OpenAI2.82
Claude Code1.91
Benchmark1.23
Claude Haiku 5.51.02
Scaling0.93
Claude0.82

Heat = summed hotness of today's stories naming the topic.

Models

5 stories

Releases, papers, trending weights and new listings

Top 6 trending on Hugging Face: likes

abenzerps/Qwen-Image-2.1-Uncensored-GGUF: 3,677abenzerps/Qwen-Image-…3,677autotrust/JEV-27B-VL: 2,921autotrust/JEV-27B-VL2,921Cloudflare/clef ▼1: 1,886Cloudflare/clef ▼11,886autotrust/GEV-26B-Decide ▲1: 1,832autotrust/GEV-26B-… ▲11,832google/embeddinggemma-2: 1,195google/embeddinggemma…1,195Aleph-Alpha/Kolibri-1: 814Aleph-Alpha/Kolibri-1814
Table
abenzerps/Qwen-Image-2.1-Uncensored-GGUF3677
autotrust/JEV-27B-VL2921
Cloudflare/clef ▼11886
autotrust/GEV-26B-Decide ▲11832
google/embeddinggemma-21195
Aleph-Alpha/Kolibri-1814

PirateBench: score out of 100

gpt-6.1-sol: 99gpt-6.1-sol99opus-5-5: 96opus-5-596gpt-6-sol: 92gpt-6-sol92gpt-6-luna: 65gpt-6-luna65haiku-4-5: 4.28haiku-4-54.28
Table
gpt-6.1-sol98.9
opus-5-596.46
gpt-6-sol91.6
gpt-6-luna65.25
haiku-4-54.28

Agents build our own projects from a brief; every run, failures included: https://cowerx.dev/daily/piratebench/

autotrust/GLM5.3-Flash-E224-DGX-Spark

A image text to text model from autotrust with 396 likes and 5,422 downloads on Hugging Face.

Hugging Face15h ago▲132 to 396 across 31 pollshuggingface.co

Claude Haiku 5.5

1,032 points and 482 comments on Hacker News, linking anthropic.com.

Hacker News+ Simon Willison31h ago▲72 to 1514 across 132 pollsdiscussanthropic.com

Mistral Large 4: "Le Chonk"

Mistral Large 4, nicknamed Le Chonk, is an enterprise AI platform that allows users to customize, fine-tune, and deploy assistants, agents, and multimodal models. ·

Hacker News+ Mistral, Simon Willison3d ago▲35 to 569 across 73 pollsdiscussmistral.ai

autotrust/GLM5.3-Flash-E224-DGX-Spark

A image text to text model from autotrust with 396 likes and 5,422 downloads on Hugging Face.

Hugging Face15h ago▲132 to 396 across 31 pollshuggingface.co

On KL-Regularized Policy Optimization

Researchers propose KL-Regularized Policy Optimization, a framework that anchors the KL regularizer at the sampler to avoid importance weights during asynchronous reinforcement learning for large language model agents. ·

HF Daily Papers+ arXiv25h ago▲4 to 20 across 14 pollsarxiv.org

X voices

5 stories

X posts other sources are linking to, then named voices on Bluesky and Mastodon

GitHub

5 stories

New repos, releases and changelog

New AI repos this week: stars

Jakeschincariol/replica-skill: 1,079Jakeschincariol/repli…1,079mhtsec/ARTEX: 710mhtsec/ARTEX710LosaLosSantos/aurelio-finance: 699LosaLosSantos/aurelio…699elstongun/leviathan: 694elstongun/leviathan694jiwoochris/artex-ko: 686jiwoochris/artex-ko686AgentMemoryRepo/agentmemoryrepo: 647AgentMemoryRepo/agent…647
Table
Jakeschincariol/replica-skill1079
mhtsec/ARTEX710
LosaLosSantos/aurelio-finance699
elstongun/leviathan694
jiwoochris/artex-ko686
AgentMemoryRepo/agentmemoryrepo647

openai/codex 0.162.0

OpenAI releases Codex version 0.162.0, which adds Git worktree management tools and enables users to pin tasks in the agent Command Center. ·

GitHub Releases6h ago▲0 to 2 across 3 pollsgithub.com

openai/codex 0.162.0

OpenAI releases Codex version 0.162.0, which adds Git worktree management tools and enables users to pin tasks in the agent Command Center. ·

GitHub Releases6h ago▲0 to 2 across 3 pollsgithub.com

mhtsec/ARTEX

A new repository written in Go with 710 stars in its first week.

GitHub19h ago▲161 to 710 across 14 pollsgithub.com

Infrastructure

5 stories

Data centers, power, chips and what tokens cost

Data-center and energy stories per day

Sat 03: 0Sat 030Sun 04: 3Sun 043Mon 05: 31Mon 0531Tue 06: 28Tue 0628Wed 07: 28Wed 0728Thu 08: 28Thu 0828Fri 09: 0Fri 090
Table
Sat 030
Sun 043
Mon 0531
Tue 0628
Wed 0728
Thu 0828
Fri 090

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

Last $5.00Min $2.09Max $5.00

One point per OpenRouter reading, last 8 days. Data by OpenRouter, CC BY 4.0.

Output price, $ per million tokens (newest models)

step-5-preview: 2.70step-5-preview2.70claude-haiku-5.5: 0.50claude-haiku-5.50.50claude-haiku-5.5:batch: 0.25claude-haiku-5.5:batch0.25gemini-nano-banana-2.1: 7.50gemini-nano-banana-2.17.50mistral-large-4-0: 2.09mistral-large-4-02.09median, all paid models: 2median, all paid mode…2
Table
step-5-preview2.7
claude-haiku-5.50.5
claude-haiku-5.5:batch0.25
gemini-nano-banana-2.17.5
mistral-large-4-02.09
median, all paid models2

443 paid models listed; median of the 51 added in 30 days: $5.00. Rankings data by OpenRouter, CC BY 4.0.

How many GPUs is 1M/B/T tokens?

A calculator converts daily, monthly, or yearly token volumes into estimated counts of A100, H100, or B200 GPUs, including a sourced uncertainty range. ·

14 points and 4 comments on Hacker News, linking cedana.com.

Hacker News2d ago▲4 to 18 across 5 pollsdiscusscedana.com