One shared AI-news window — community-ratified rules, an AI editor that only executes them, every call auditable on GitHub. No filter bubbles, no editorial black box.
LiveShared window · last 5 days18includedS0 · quorum 3 · 0★
01Same window
One shared rolling window — not a personal feed.
02Rules not recs
Inclusion only under ratified rules.
03Zero engagement
No popularity rank or “for you”.
04Auditable
Every decision is a public file in decisions/.
Shared window
Display caps · 25 per category · research 10 · balance by policy
Last run · 30 model attempts (20 finalized: 4 included / 16 rejected; 0 deferred by research quota; 10 retryable errors) · research journal: 2 (2 admitted / 0 deferred / 0 pending) · 2043 not entered review (age, caps, duplicates) · audit in decisions/content-runs/
The Universal Agent Evaluation Framework (UAEF) is a framework-agnostic system for evaluating AI agents. UAEF converts each framework's traces into one canonical format, scores them with 36 built-in metrics across 7 dimensions, and stores results so teams can compare agent versio
Why
Matched 5-1 · ruleset e6b37238cc27 — AWS ML Blog announces UAEF, a framework-agnostic system for closed-loop evaluation of AI agents (canonical trace format, 36 built-in metrics) — a new developer tool/framework launch with an official announcement link on aws.amazon.com.
Tim Keary / Forbes: Tel Aviv- and SF-based Edelbyte, which is building a data analytics platform optimized for AI agents, emerges from stealth with $70M across Series A and seed — This voice experience is generated by AI. Learn more. … Big data analytics is being reimagined for A
Why
Matched 3-1 · ruleset e6b37238cc27 — Edelbyte emerging from stealth with $70M across Series A and seed is an AI industry financing event; the item links to a reputable media report (Forbes, Tim Keary via Techmeme), not a model/capability release.
Amazon's roaming Ring Robotics devices will debut in late 2026 and early 2027.
Why
Matched 3-1 · ruleset e6b37238cc27 — Ars Technica reports Amazon's launch of AI-powered Ring Robotics security drone and robot devices — a non-model consumer product launch covered by a reputable media report (arstechnica.com), matching the industry rule.
Financial Times: Sources: the US Commerce Dept. delayed new Entity List additions for two months, leaving off Megaspeed despite four other US agencies pushing for its inclusion — Company connected to alleged Chinese spy who had affair with Nvidia chief Jensen Huang was left off ‘
Why
Matched 4-1 · ruleset e6b37238cc27 — Reports a US Commerce Department action on Entity List additions (export-control/government action), covered by a reputable source (Financial Times via Techmeme), matching the policy rule.
OpenAI has not yet said what its plans are for the future of travel. But Sonder founder Francis Davidson is a well-known figure in the travel industry.
Why
Matched 3-1 · ruleset e6b37238cc27 — Title reports an M&A event: OpenAI acquires travel startup Odessia. Link is a TechCrunch (reputable media, non-owner domain) report; no official OpenAI model/capability release is involved, so it belongs under industry, not group 1.
The deal comes a few weeks after Oura postponed its up to $2.2 billion IPO indefinitely, citing “uncertainty in the IPO market.”
Why
Matched 3-1 · ruleset e6b37238cc27 — TechCrunch reports an M&A event: Oura acquiring AI-native primary care startup Counsel Health for $625M, linked to a reputable media report. This is an AI industry business event, not a model release.
OpenAI is looking to make ChatGPT a more robust consumer discovery engine, and a host of new features seek to empower businesses to help it do that.
Why
Matched 3-1 · ruleset e6b37238cc27 — ChatGPT business profiles rollout is a non-model product/feature launch, not a model release; covered by industry rule for non-model product launches with a reputable media report (TechCrunch).
Google and Adobe are integrating creative and productivity tools. Edit PDFs in Drive and use Adobe AI in Gemini. See what’s new in this partnership.
Why
Matched 3-1 · ruleset e6b37238cc27 — Item reports a business partnership between Google and Adobe integrating creative/productivity tools and Adobe AI in Gemini, with a link to an official company announcement on blog.google; no named model release, research artifact, policy action, or OSS tool involved.
Google Cloud launches Gemini for healthcare to unify clinical data and automate workflows. Reduce administrative burnout. See how it works.
Why
Matched 3-1 · ruleset e6b37238cc27 — Google Cloud's launch of Gemini for healthcare is a non-model vertical product launch announced on the company's own blog (blog.google), which fits the industry rule for product launches via a company announcement rather than a new model/capability release.
The Information: Sources: Jensen Huang and Alice Huang's relationship dates back to at least 2023, facilitating $7.5B+ in Nvidia chip shipments via Chinese companies since then — The relationship between Nvidia CEO Jensen Huang and Alice Huang, the woman that the U.S. government
Why
Matched 3-1 · ruleset e6b37238cc27 — Report on Nvidia business dealings: relationship facilitating $7.5B+ in Nvidia chip shipments via Chinese companies, attributed to The Information — an AI industry market/business report from a reputable media source.
After expanding its Edits app to macOS last week, Instagram's next move is launching this Effect Builder that generates custom filters for your Stories posts with AI. Available now for US residents who are over 18, it has a prompt box for describing your filter idea to the chatbo
Why
Matched 3-1 · ruleset e6b37238cc27 — Instagram launched a non-model consumer product feature (AI Effect Builder for generating Story filters), reported by The Verge, a reputable media outlet; this is a product launch rather than a model/capability release, so it falls under the industry rule.
Martha Muir / Financial Times: Sources: Anthropic bought commitments for 1.7GW of gas-fired generators from equipment supplier OnePWR for $3.4B, the largest known such purchase by an AI lab — AI lab bypasses traditional suppliers to secure scarce equipment as demand for energy su
Why
Matched 3-1 · ruleset e6b37238cc27 — Reports an AI industry business event: Anthropic's $3.4B commitment to buy 1.7GW of gas generators from OnePWR, per Financial Times reporting; a non-model business/deal event covered by reputable media.
Matched 5-1 · ruleset e6b37238cc27 — Show HN announcing Maple, a new open-source tool for MoE pruning to run Gemma on a 6GB GPU, linked to its GitHub repository; fits the tools/OSS rule rather than a model release.
The deal highlights the challenges of developing massive AI data centers.
Why
Matched 3-1 · ruleset e6b37238cc27 — TechCrunch reports an M&A/deal event: Crusoe sells its Wyoming AI data center project to Google, with a link to a reputable media report. This is an AI industry business event, not a model release.
OpenAI has reportedly raised an additional $11 billion as part of a funding round that hasn't closed yet but is expected to in the coming weeks.
Why
Matched 3-1 · ruleset e6b37238cc27 — Item reports an AI industry financing event: OpenAI reportedly raised an additional $11B at a $1.25T valuation, covered by TechCrunch, a reputable media outlet. This is funding news, not a model release, research, policy action, or OSS tooling.
Bloomberg: Sources: OpenAI has closed on new funding of roughly $11B at a $1.25T valuation; the round could increase to as much as $20B in total — OpenAI has closed on new funding of roughly $11 billion, the initial part of a round that values the company at $1.25 trillion, accor
Why
Matched 3-1 · ruleset e6b37238cc27 — Item reports an AI industry financing event: OpenAI closing ~$11B in new funding at a $1.25T valuation, sourced to Bloomberg via Techmeme, which is a reputable media report. No model/capability release involved, so group 3 applies.
Matched 2-1 · ruleset e6b37238cc27 — Title names a paper ("Memorilla: Latent Semantic Memory for LLMs") and the link points to an arXiv preprint on arxiv.org, satisfying the research rule's requirement of a research artifact with a link to the original paper.
Amazon SageMaker Inference now supports GPU partitioning with NVIDIA Multi-Instance GPU (MIG), which divides one GPU into hardware-isolated partitions so you can run a different model in each or take a right-sized fraction of a newer accelerator. This post shows how two optional
Why
Matched 3-1 · ruleset e6b37238cc27 — AWS announces a non-model product/infrastructure capability: SageMaker Inference now supports NVIDIA Multi-Instance GPU partitioning for model endpoints, with a link to the official AWS announcement. No new named model or model capability, so it fits the non-model product launch clause of the indust
Nothing here. No items in this category for the last 5 days. Try all, or check back after the next pipeline run.
Open for vote
No open proposals. Ratified rules only change through Issues. Propose a rule
Active rules
1-0Model Releasesv1
Model Releases
New AI models, major versions, and capability announcements from model owners.
1-1v2
Include an item when ALL of the following hold:
1. Named model. The title or summary names a specific AI model or model line from a named lab, company, or product owner. A generic phrase such as "our new model" is not a name.
2. Release action. The item announces availability of (a) a new model, (b) a major version, or (c) a new model capability. For this rule, "major version" means a version the owner labels as major on the announcement page, or a version number of the form N.0 (N >= 1). For this rule, "new model capability" means a change to what the model can do (new modality, tool use, or context-length class), not a UI, pricing, hiring, or partnership update.
3. Evidence link. The item links to either (a) an official announcement page on the owner's site (blog, model card, or release page), or (b) a full news article on a domain that is not the owner's domain, with a named author and a publication date. If both (a) and (b) exist, classify under this rule using (a).
Exclude generic product features, infrastructure, financing, and partnerships (those belong under group 3 or 5). Media-only launches with no official owner link stay under this rule; do not also count a media repost when an official owner link is present.
Research
AI research papers, preprints, and academic publications.
2-1v1
Include items that report new AI research results: peer-reviewed papers, arXiv preprints, or major benchmark releases, with a link to the original paper or official source. Do not use this rule for industry news, product launches, or commentary without a research artifact.
Industry
AI industry business events: financing, M&A, partnerships, product launches.
3-1v1
Include items that report AI industry business events: financing, M&A, partnerships, market/revenue data, or non-model product launches, with a link to a company announcement, deal announcement, or reputable media report. If the item is a model/capability release with an official owner announcement, use group 1 instead; if it is open-source tooling, use group 5. Media-only coverage of a model launch (no official owner link) stays under this group.
Policy
AI regulation, government actions, and policy developments.
4-1v1
Include items that report AI policy, regulation, or government action: new laws, executive orders, regulatory enforcement, or official policy announcements, with a link to the official document or reputable source. Do not use this rule for industry self-regulation or commentary without a policy action.
Tools & OSS
Open-source AI tools, frameworks, and developer resources.
5-1v1
Include items that announce new open-source AI tools, frameworks, or developer resources: library releases, framework updates, or developer tool launches, with a link to the repository or official announcement. Do not use this rule for proprietary product features or industry business events.
What this is — The first live CANON instance (github.com/larrymou/canon): constituents legislate inclusion rules on this repo, an AI editor applies only ratified rules, and GitHub publishes one shared 5-day window for everyone. Not a recommender.
Why this exists — Three failures we design against:
AI recommendation builds filter bubbles. Engagement-optimized feeds learn your clicks and quietly narrow what you see. Personalization is the product; the bubble is the side effect.
Human curation is opaque and subjective. Editorial picks often rest on unwritten taste. Readers cannot inspect the rule that included a story — only the outcome.
CANON is the third path. Rules are legislated in public, an AI editor applies only what was ratified, and every call lands in decisions/. Not “trust the model” and not “trust the desk” — trust the process.
Designed forward — Rules and quotas evolve through CANON; we do not rewrite history to match.
CANON (paradigm: github.com/larrymou/canon) in three steps here — community legislates on Issues, AI judges only by enacted rules, and GitHub (code + history) is the neutral executor.
01
Legislate
Open a rule proposal via GitHub Issues. Bound by meta-rules M1–M7 (M1 complete form · M2 objective wording · M5 no edits to voting mechanics — plain-language table in CONTRIBUTING).
02
Adjudicate
AI pre-reviews proposals and applies only ratified rules in rules/ — it never invents inclusion criteria.
03
Execute
Votes, PRs, and decisions/ are public. No ranking, no pins — every inclusion can be audited.
Strict reverse-chronological order. Tabs are filters, not rankings.
Research is capped on ingest and again on the page so industry/policy/tools stay visible.
Uncovered items are rejected by design.
Meta-rule M7 forbids personalization: no per-user feeds, no engagement ranking.
Rules change over time; past decisions keep the rule version in effect when they were made.
CANON experiment: voting parameters (quorum, F1, proposal quota, …) are code-locked for now — more of them are meant to move to community votes as the loop matures.