KryptonAi by Alexandru Dan

🚀 Breaking AI & Emerging Tech First
🤖 Real-time #AI, #Web3, #Quantum, #FutureTech
⚡ Trusted by innovators, built for visionaries
🌍 Curated by KryptonAI

[KryptonAi by Alexandru Dan @KryptonAi]

@Umesh__digital I dont even consider Windows. Mac is the best for coding!

[Vaibhav Sisinty @VaibhavSisinty]

Baidu just open-sourced an OCR model that reads entire 40-page documents in one shot.

It's called Unlimited-OCR. 3 billion parameters but only 500 million active during inference. Runs 100% locally on your machine.

Why this matters: traditional OCR tools chop documents page by page. Tables that span two pages break. Reading order gets lost. Cross-page context disappears.

Unlimited-OCR processes the whole document at once. 32K context window. Text, formulas, tables, reading order all preserved across pages.

Output comes out as clean structured Markdown.

→ 93% accuracy on the standard benchmark. +6 points over the baseline.
→ Error rate stays below 0.11 even past 40 pages.
→ Multilingual out of the box.
→ 2.12 million downloads on Hugging Face last month. 14,600 GitHub
stars.

For context: Amazon Textract, Google Cloud Vision, and Azure Document Intelligence all charge per page. This runs locally for free.

[rari @0xwhrrari]

Andrew Ng just released a 2-hour course

On building agentic skills from scratch with Anthropic:

00:00 - How to build agent skills with Claude
22:32 - Claude pre-built skills for AI agents
41:07 - Agentic skills vs tools, MCP, and subagents
01:06:06 - Skills for long-running agents

This 2-hour watch can replace 10 paid courses on building agents

Taught with Anthropic themselves

Bookmark and watch it tonight

Then read the article below

[Jean-Rémi King @JeanRemiKing]

We’re happy to announce 2 releases today:
- 🧠Brain2qwerty v1 is published at @NatureNeuro
- 🚀 Brain2Qwerty v2 is now publicly released

Explore how we decode sentences from non-invasive brain recordings: https://t.co/IdR6gK2hcd

Links:
📄v1 Nature Neuro: https://t.co/wnRjc9W9gI
📑v2 Meta preprint: https://t.co/oSfLOQFcvg
💻Code: https://t.co/Xbe0XWfWQL
📊Data: https://t.co/SCBbs4AhTg
📝Blog: https://t.co/15RvsAaXlH
🧵Thread: https://t.co/d8FJrVyDut

[Rowan Cheung @rowancheung]

There's a startup trying to build data centers in the ocean.

And it's INCREDIBLY fascinating:

Mass consumption of electricity and water is a growing bottleneck for data centers.

So by moving offshore, it eliminates both problems -- the ocean provides unlimited cooling, and the waves provide unlimited power.

There are also no engines, so the data centers drive themselves to their destination by using the shape of their hull to propel through waves.

Called Panthalassa.

[KryptonAi by Alexandru Dan @KryptonAi]

Interesting!

[Soham Mehta @sohmehta]

Today we're launching Goose Ads in Claude.

This is a skill /goose-ads that lets anyone make high-performing ad creatives directly in Claude, Claude Code, Cowork, or Codex.

It finds the ads companies are already paying real money to run and remakes them for your brand. Accurate logo, messaging, and assets. One prompt.

Here's how it works:

1. Install: npx gooseworks install --all

2. Run: /goose-ads create ads for my brand [your-website]

3. Pick the templates you like

That's it. Winning ad creative in minutes, inside Claude.

But this is just the start.

We open-sourced 100+ growth skills that some of the fastest-growing startups run every day. Ads, content, competitor research, GTM, SEO, all of it.

Comment Goose and I'll DM you all 100+. For free

[KryptonAi by Alexandru Dan @KryptonAi]

I like this!

[Ning @totheagi]

made this game with GLM-5.2 (claude code) running on my 32x RTX 4090s

[KryptonAi by Alexandru Dan @KryptonAi]

@svpino :))

[Boris Cherny @bcherny]

Cool way to use Claude Code: deciphering Linear A, a 3500 year old written language from Crete

https://t.co/Aqd4ZG7Cum

Hope this holds up in peer review! 🤞

[KryptonAi by Alexandru Dan @KryptonAi]

Training loss constrains predictions on the training distribution; it does not uniquely determine behavior on unseen prompts. A model can achieve zero training loss while producing outputs absent from its training set. Therefore, novel outputs do not mathematically imply high SFT or RL loss.

[KryptonAi by Alexandru Dan @KryptonAi]

@paradite_ That is false.

[KryptonAi by Alexandru Dan @KryptonAi]

That is false.

[Zhu Liang @paradite_]

i’m really surprised that people don’t see this.

It’s mathematically true that llms can’t come up with novel ideas, because the whole point of training is to reduce loss, gain rewards so that the model adhere to rules and ground truth.

if you have a model that can come up with novel ideas, it must have high loss during sft or rl.

[KryptonAi by Alexandru Dan @KryptonAi]

Panthalassa, a startup from Portland, just raised $140 MILLION. They are putting DATA CENTERS in the ocean now.

[KryptonAi by Alexandru Dan @KryptonAi]

This is happening inside AI brain!

[Airbazoo @airbazoo]

THIS IS NOT AI, This Is BENI!

Beni is an autonomous two wheeled self balancing camera robot who’s ready to be your next camera assistant!

Beni can automatically track and follow you, has built in storage for recording, and can traverse over various terrains, and even jump and flip in a variety of directions.

Beni might be the future of filmmaking and capturing authentic moments. Beni will be available on Kickstarter from July 8 - August 21, will you be getting one? 👀

Never Give Up Keep Creating ♥️

#robot #camera #camerarobot

[KryptonAi by Alexandru Dan @KryptonAi]

Interesting ideea!

[alphaXiv @askalphaxiv]

“Self-Evolving Multi-Agent Systems via Decentralized Memory”

Multi-agent systems usually share one memory pool, but that makes agents converge toward the same behavior and lose useful specialization.

This paper gives each agent its own memory, split into reuse and exploration pools, so agents can learn from past collaboration without collapsing into one shared strategy.

This paper's DecentMem gets O(log T) regret, up to 23.8% better accuracy than centralized memory, and up to 49% fewer tokens.

[KryptonAi by Alexandru Dan @KryptonAi]

This is a good joke!

[Alexander Knigge @AlexanderKnigge]

Okay @MistralAI Le Chaton Fat model is officially insane...

Their CLI handles native LaTeX rendering and the 30T MoE with 256 experts absolutely CRUSHES differential geometry

We are so back

[KryptonAi by Alexandru Dan @KryptonAi]

Current AI systems are not the ideal machines for determining truth. They are fundamentally engines that create "highly probable statements" – but the probability is not that of being true but of being said.

[Anthony Aguirre @AnthonyNAguirre]

https://t.co/acyvgIN1OE

[KryptonAi by Alexandru Dan @KryptonAi]

Satya’s take on the "cognitive loop" is a must-read for the new economy.

[Satya Nadella @satyanadella]

https://t.co/vLmiBKTtX3