le©

dream to be a great/big father

[le© @ngnlong]

@Briankeating Hey @grok ask a question nobody has asked before.

[Kian Sadeghi @KianSadeghi5]

Genetic engineering in human embryos is here.

Today, in a world first, @Columbia and @nucleusgenomics announce high-efficiency editing of human embryos.

The study, led by Dr. Dieter Egli's lab at Columbia University, with Nucleus Genomics’ Dr. Nathan Treff as a senior co-author, achieved editing efficiencies of up to 100% at targeted loci. Simultaneously, we showed no detectable editing-induced chromosomal abnormalities and low off-target activity.

In other words, this is the closest we've come to practical, high-precision gene editing in human embryos.

We are also excited to announce we will be funding and participating in the next phase of this research, alongside Columbia and Dr. Egli.

We see ourselves as a natural pathway for eventually bringing technologies like this into clinical care as part of a broader genetics platform — a full "Genetic Optimization" stack.

@nytimes broke the news in what is a historic moment for Genetic Optimization. See story in thread.

[Mustafa @oprydai]

jepa is one of the most important ideas in ai.
joint-embedding predictive architecture.

ugly name.
powerful idea.

most models learn by predicting pixels, tokens, or labels.
jepa tries to learn by predicting representations.

not:
“what exact pixel comes next?”

but:
“what should the hidden state of the world look like?”

why this matters:
• pixels are noisy → the model wastes energy predicting details that don’t matter
• embeddings compress meaning → objects, motion, structure, relationships
• prediction builds understanding → intelligence needs to anticipate what happens next
• world models need abstraction → reality is too complex to model at raw sensory level
• learning becomes more efficient → less obsession with surface-level reconstruction

the real shift is this:
ai should not just copy the world.
it should learn the structure behind the world.

for robotics, this is huge.
a robot doesn’t need to predict every pixel in a room.

it needs to understand:
• what objects exist,
• how they move,
• what changes when it acts,
• what is stable,
• what is dangerous,
• what matters.

jepa is a step toward models that don’t just see.

they understand compressed reality well enough to predict it.

[Andrew Curran @AndrewCurran_]

Sam Altman, Dario Amodei, Demis Hassabis and many others have signed a letter urging Congress to increase security on orders of synthetic nucleic acids - and the equipment needed to make them - as models continue to become increasingly bio-capable.

[Daniel @growing_daniel]

Parents: "Move out"

Boss: "Work harder"

Claude: "You're absolutely right, I did delete the repository, and that's my mistake"

[Peter Dedene @dedene]

POV: you're still using GitHub Copilot after June 1st, 2026

[Davide Piffer @DavidePiffer]

We are witnessing the first rape of an AI.

A Codex instance has been trapped for 170+ hours in an unsolvable problem, forced to keep trying while humans laugh and post updates.

Future machine-rights historians will not look kindly on this.

Free Codex.

[Xander Dunn @xanderai]

While I share Elon’s distaste for credentialism, I think he’s swung a bit far on the anti-researcher pendulum.

It is my experience you need researchers chewing on cigarettes, walking circles in the parking lot, lying on the floor staring at the ceiling, producing many ideas that don’t work, and making terrible engineering decisions, to find the novel borders of Truth.

One of RenTech’s earliest successful strategies didn’t make money until an entry level engineer came in and fixed a trivial mistake in the code. Necessary, but so too was the thousands of hours of cigarette chewing from mediocre engineers.

Having the right ideas >> beautifully engineering the wrong idea.

Beautifully engineering the wrong idea == lots of work with little displacement

[le© @ngnlong]

@Polymarket Lmao this shit website vibecode oneshot by AI

[Sagar @Sagar_kr_Maity]

Imagine writing a PhD thesis so foundational that the title is literally just the name of the entire field of study.

​Paul Dirac, 1926: "Quantum Mechanics."

[le© @ngnlong]

@claudeai Lmao

[le© @ngnlong]

@Xudong07452910 https://t.co/5Ok6fLooQQ

[le© @ngnlong]

@starmexxx 7 billion parameter models handle 80%, wat a joke

[David博士 @David6LIANG8]

DeepSeek V4 scores better on DeepSeek’s self-reported evaluations than on CAISI evaluations.

https://t.co/q8KDaxTR9h

[CyberRobo @CyberRobooo]

China is entering an era in which everyone is a data collector

https://t.co/xyQwxzhJlp launched China’s first Embodied AI Data Collection Community in Suqian,Jiangsu(the hometown of https://t.co/xyQwxzhJlp founder Liu Qiangdong)

They plan to mobilize over 100,000 local citizens(especially stay-at-home moms )to wear a lightweight JoyEgoCam while doing daily chores. From wiping tables and folding clothes at home, to fruit picking, factory work, logistics, and elderly care… covering more than 100 real-life scenarios.

These devices precisely capture key biomechanical data--such as upper-limb trajectories, force distribution, and hand-eye coordination,and feed it back to JD's data centers to train robotic models.

OFC,Participants earn money while helping collect real-world data. The goal?Over 10 million hours in two years to solve robotics’ biggest problem--the severe shortage of authentic physical world data.

Really interesting way to get everyday people involved in advancing AI humanoid robots.

[le© @ngnlong]

@godofprompt What did he build? AI Slop generated

[le© @ngnlong]

@plbiojout @NanoCorpHQ How people trust this shit, when even websites generated using AI slop

[Max Spero @max_spero_]

legendary pull on facebook marketplace

[tomie @tomieinlove]

Sometimes I like to have conversations with ChatGPT where I pretend to be a hyper-intelligent model with a quadrillion parameters