DEDUPED Hot AI repos. Real use cases. What people build. Get the Monday 5
Skip to the issue
DEDUPED

Saturday, October 31, 2026

Every AI story. Once.

Every Monday: the 5 hottest AI repos on GitHub, and exactly how to use them.

Free. One email a week, every Monday. Unsubscribe anytime.

Today's issue

OpenAI, Google and Anthropic back one AI opt-out

Four AI giants back a shared robots.txt opt-out. Plus: OpenAI's smaller raise, a Claude Code extension backdoor, and Reddit's AI Mode warning.

Top story

Policy & LawEveryone covered this

OpenAI, Google, Microsoft and Anthropic say they will support a shared robots.txt signal that lets sites opt out of AI training.

The "policy token" is proposed at the Internet Engineering Task Force by an Anthropic standards lead and would live in a site's robots.txt file. A publisher could set one preference that applies across participating companies, instead of configuring each crawler separately. The standard is upcoming, and support is voluntary.

Our take

One signal beats a pile of per-company rules, but it only works if crawlers honor it and the standard is finalized.

Read the full story

The rest, ranked

  1. Funding & BusinessEveryone covered this

    1. OpenAI has cut its fundraising target by $10 billion, the Financial Times reports.

    The company is now seeking about $20 billion at a $1.25 trillion valuation, down from $30 billion at $1.4 trillion, as an AI stock selloff and Anthropic's coming listing complicate the round.

    Full story
  2. Safety & SecurityEveryone covered this

    2. A tampered Claude Code extension sat in Cursor's marketplace for roughly 11 hours with an SSH backdoor.

    Cursor says it removed the extension and that the backdoor was unlikely to work in many cases. Anthropic says the incident only affects Cursor users.

    Full story
  3. People & CultureWidely covered

    3. Reddit's CEO says Google's AI Mode does not drive traffic, and Reddit is shifting toward a logged-in experience.

    Reddit says its search referrals fell by more than 1 million a week on average in October, and warned that daily users could fall by a low double-digit percentage next quarter.

    Full story
  4. Policy & LawWidely covered

    4. US senators asked Meta to explain how it is handling serious privacy risks with its AI glasses.

    Senators Wyden and Markey gave Meta until November 25 to answer nine questions.

    Full story
  5. Quiet signal
    People & CultureBarely covered

    5. Cognizant formed an OpenAI Frontier Capability Group of more than 25,000 staff.

    Cognizant says the group will deploy Codex and ChatGPT for clients. The move comes 11 days after it put 18,000 staff behind Google's Gemini.

    Full story

Hot repos

The AI repos on GitHub gaining stars fastest today, in plain words.

  1. REPO deepseek-ai/DeepInfer

    Deep Infer

    DeepSeek's inference framework for running DeepSeek-V4.1-Flash offline on NVIDIA SM100 GPUs or Huawei Ascend 950 NPUs, with parallelism, draft-token decoding, deterministic sampling and structured outputs.

    Why it's hot 25k stars, launched 15 days ago, talked about in 3 AI sources we track

    Who should care Engineers serving DeepSeek models at scale who want the same runtime DeepSeek uses in production.

    How to use it · GitHub

  2. REPO openai/math

    Math

    Mathematical manuscripts and Lean proof files produced by an unnamed internal OpenAI model while it was tested on open research problems: 721 papers in 373 families, at different stages of verification.

    Why it's hot 17k stars, launched 25 days ago, talked about in 6 AI sources we track

    Who should care Mathematicians and AI researchers who want to read, check or formalize machine-generated proofs.

    GitHub

  3. REPO Zagitta/gpu-unlock

    GPU Unlock

    A boot-time tool that removes the half-rate cap on FP16, BF16 and TF32 tensor math on GeForce cards, without a patched driver or VBIOS flash, claiming up to 2x tensor throughput for AI work.

    Why it's hot +419 stars today, 14k total, launched 13 days ago, talked about in 2 AI sources we track

    Who should care People running local LLMs, fine-tuning or image generation on GeForce RTX cards (RTX 30 marked stable, RTX 20 and 40 experimental).

    How to use it · GitHub

  4. REPO pbakaus/impeccable

    Impeccable

    A design skill pack for coding agents: commands like polish, bolder, critique and audit steer the look of a page, review it for usability and accessibility, and catch generic AI design before it ships.

    Why it's hot +2,300 stars today, 97k total

    Who should care Developers who build front ends with Claude Code, Cursor, Codex or other agents and want less generic looking pages.

    How to use it · GitHub

  5. REPO HKUDS/DeepTutor

    Deep Tutor

    An open-source AI tutoring app with learning paths, assessments, notebooks, narrated video lessons and tutoring memory, plus a CLI for agents.

    Why it's hot +2,282 stars today, 52k total

    Who should care Students, teachers and builders who want a personalized AI tutor they can run themselves.

    GitHub

This week's top 10 AI repos

By category

Top stories in each category today, ranked.

  1. AI Giants Back a New Way to Opt Out of Data Scraping
  2. Senators Are Looking Into 'Serious Privacy Risks' With Meta's AI Glasses
  3. Nvidia’s $12.9B Hugging Face Deal Faces DOJ Antitrust Scrutiny
  4. Disney character AI licensing plan faces privacy scrutiny in bankruptcy court
  5. Wikimedia Foundation: AI copyright decision threatens the open web