AI, distilled & clear.
I read the daily AI firehose for you — fact-checked and explained in plain language — every piece a visual explainer with motion, interactives, analogies and a podcast, in EN / 中 / 日.
Per article, monthly or yearly — your call.
Founding seats are sold out — 100 members locked the lifetime price. Benefits only grow.
Card / Alipay / WeChat Pay · 7-day refund | $59 once we reach 100 paying members — your price locks for life
Subscribe yearlyBilled in USD via Stripe · Alipay / WeChat Pay / card / Apple Pay · CNY is an FX estimate.
Team · 5 seats $199/yr · email us →
Dozens of AI headlines a day — fewer than three deserve your time.
X, papers, launches, newsletters — a flood. It all gets scanned so only what's worth reading is left.
Press releases oversell; second-hand takes are everywhere. Every piece is cross-checked, and vendor content gets a ⚑ flag.
Papers and jargon push people away. Analogies, motion and interactives break it down until it clicks.
Not a wall of text — an article you can play with.
Faster at the same power — higher training and inference throughput.
Far less power at the same compute — electricity and cooling are the biggest datacenter cost.
- One-line analogy — get transistor stacking in a second
- Motion + interactives — click through the mechanism yourself
- Key data tables — vendor claims vs. measurements, kept apart
- Podcast version — the day's insight on your commute
- EN · 中 · 日 — three languages, updated together
Three steps, every day.
Scan everything: official blogs, papers, launches, first-hand posts, newsletters.
Cross-check sources, separate claims from measurements, flag vendor spin with ⚑.
Analogies + motion + interactives, with a podcast, in three languages.
News is free. Insight is for members.
Product launches and business moves are always free in full. 🔒 tool tutorials, deep dives and research explainers: free preview up front, full text is for members.
Free keeps you informed — the why lives behind the wall.
| Free reader | Member | |
|---|---|---|
| Product news · business | ✓ Full text | ✓ Full text |
| Tool tutorials · deep dives · research | Free preview | ✓ Full text + entire archive |
| Podcast audio | News episodes | ✓ Everything · private feed |
| Full-text Markdown export (Obsidian / Notion) | — | ✓ One-click copy / download |
What you might ask.
What can I read without paying?›
Why a paywall instead of ads?›
Does the wall hurt search and discovery?›
How does per-article buying work?›
Refunds?›
How do I sign in? Paid but can't get back in?›
How do I cancel?›
Every piece follows four baselines.
Every piece links its sources and separates claims from measurements. Vendor content gets flagged.
Not just what launched: why it matters, and where it breaks down.
Analogies, animation and interactives turn dense papers into things you can actually grasp.
Warm, plain, conversational. A smart friend explaining — not a brand broadcasting.
The steadier the membership income, the less this site has to please anyone — your money also buys its independence.About the site →
Reading now
All →Sam Altman on past mistakes, distillation, and the three things AGI still needs
He also admits that in 2019, the entire field expected AI to upend the economy—and it didn't.
LayerX finds a new visual fraud attack that tricks AI browser assistants
A custom font and a few lines of CSS are enough—no JavaScript, and no browser vulnerability required.
Replit launches Replit Design: one click for another layout, and a mockup that becomes a working site
Canvas and click-to-edit are free; turning a design into an App starts at Core for $25 a month.
Two API settings, not a new model: OpenAI reports GPT-5.6 Sol jumping from 13.3% to 38.3% on ARC-AGI-3
The model stayed the same; the harness did not—retained private reasoning plus compaction instead of deletion cut output tokens per game to about one-sixth.
Magnific's 38-page handbook for image and video prompts that no longer work on frontier models
Copy-ready prompts, banned-word lists, and camera rules are all there—but the model rankings come from Magnific's own team, with no disclosed test method.
Claude Mythos finds pure mathematical flaws in encryption schemes, matching elite cryptographers
Neither result threatens anything in production: HAWK is not deployed yet, and the AES work hit only a seven-round reduced version, not the full ten-round standard.
Tomorrow morning, dozens more things will happen in AI.
The few that matter will be explained here.