Vikram Dutt

[Ethan Mollick @emollick]

We already have superintelligences that have taken over the planet: we call them organizations.

[Abhishek Nagaraj @abhishekn]

I honestly haven’t seen any literature linking distributed knowledge in organizations as a form of IP protection. But it seems like it *must * be right. In which case there are two implications - (a) please point me to work on this spirit that I don’t know about and (b) if you are an enterprising grad student in innovation econ/strategy this topic is gold!

[Guillermo Rauch @rauchg]

👨‍💻 https://t.co/e8wWoxxSi4

[Ilir Aliu @IlirAliu_]

Do you know how much factory automation still dies on soft materials, like cable routing, and part variation???

This week, a Video-Action Model from @mimicrobotics is already running those exact jobs on @AudiOfficial lines after minutes (!) of demonstration data. Yes, you read that right. Not weeks, around 30 min!

FLUX-mimic (built with @bfl_ai) treats robot control as visual prediction.

On soft-body kitting, it hits 95% success straight out of the box.

No heavy task-specific fine-tuning!

Traditional approaches needed faaaaar more data and still landed lower.

If you work or oversee real production tasks, you know that today it's simply not possible to automate those tasks like door seals, flexible insertion, parts kitting.

The kind of work most conventional systems still struggle with or simply skip.

Mimics approach? It runs on a single edge GPU, and deployment data feeds back into the model.

Integration time collapses because the system already understands how the world moves.

If you are still budgeting flexible automation the old way (months of programming + massive demo collections), have a look at this. Seriously.

Worth a closer look if soft materials or high-mix assembly sit on your roadmap.

[Jeremy Jordan @jeremyjordan]

last weekend i set up some automation so that chatgpt could generate a meal plan and workout plan for a rolling three day window. chatgpt work (cloud) has daily automation to generate the plan, uses a custom MCP server, hosted on cloudflare, to track the generated plans. i have a dashboard i can access as a progressive web app, and i also built a custom TRMNL plugin so i can display the plan on a nice e-ink display. here's a diagram of the setup courtesy of codex + tldraw offline.

[Alvaro Cintas @dr_cintas]

your AI agent can watch any video now - paste a URL and it sees every frame, hears every word, all for free 🤯

bradautomates/claude-video gives Claude the ability to watch YouTube, Loom, TikTok, local files - anything yt-dlp supports

what people actually use it for:

→ analyze a competitor launch - what hook, what visuals, what structure
→ debug from a screen recording - Claude reads the exact frame where it breaks
→ summarize a 49-min talk in 30 seconds with frame-accurate timestamps
→ strip the hype from product videos - "what's actually new, skip the pitch"

the mechanism: yt-dlp pulls free captions first (zero cost). ffmpeg extracts frames at scene-aware intervals - not uniform sampling, so you don't waste tokens on 12 identical frames of the same slide. Claude reads every frame as an image with timestamp markers. Groq Whisper only kicks in when a video has no caption track

how to set up (3 min):

> claude code: /plugin marketplace add bradautomates/claude-video then /plugin install watch@claude-video
> or npx skills add bradautomates/claude-video -g for codex, cursor, gemini cli
> dependencies auto-install on macOS via brew

two caveats: free captions cover most but not all videos. past 10 min use --start/--end for focused sections or the token-burner mode for full coverage

your buddy still watches every tutorial at 2x speed taking manual notes. you paste a URL and your agent extracts the substance in seconds for $0

[Kirk Borne @KirkDBorne]

Tensors — The Geometric Tool That Solved Einstein’s General Relativity Problem: https://t.co/3TsQmYcAmz
++
Tensor Decompositions for Data Science [and Computational Science]: https://t.co/WofrhrRiGo
— Extensive background materials in linear algebra, optimization, probability, & statistics are included as appendices in this book.

[Dominik Tornow @DominikTornow]

Can you one-shot durable executions over lunch?

https://t.co/SGXPaajnQ3

[Andrew D. Huberman, Ph.D. @hubermanlab]

Once you embrace that attention and the neural circuits for holding attention are plastic/malleable, you’re on your way to better focus.

Learn to ignore/not engage in meaningless drama or numbing out. Or get passed by.

Bite down into some hard work/learning for hours a day.

[Taelin @VictorTaelin]

Update: just added a command to expand a compressed memory into two larger parts. This lets the agent navigate the memory tree, allowing it to seek and recall information from ages ago, with arbitrary precision. If you cloned it already, pls update!!

[Aryaman Arora @aryaman2020]

the main thing which distinguishes interp from the rest of AI research right now is that it is still limited by quality of ideas and not by compute. as a result i’m skeptical of 10xing progress claims!

[John Maeda @johnmaeda]

RL-ING YOURSELF: The LLM can write better than you, by a standard which others are measured. But you write better than the LLM, by a standard which you measure yourself by. To remain yourself, it is important to write by yourself by. At least if you care to be. Yourself. And not just. The Others.

Because writing wrong, and actively correcting yourself, is how you learn. It is the path to becoming your new self. It is how you keep growing. It is how you remain. Yourself.

—JM updated a year later

[Tech with Mak @techNmak]

In 1683, a Swiss mathematician named Jacob Bernoulli asked a simple question about money.

If you deposit $1 at 100% annual interest, what happens if you compound it more and more frequently?

Once a year → $2.00
Twice a year → $2.25
Monthly → $2.613...
Daily → $2.714...

The more frequently you compound, the higher it goes, but it never grows without limit. It converges.

Bernoulli proved that limit exists:

e = lim (1 + 1/n)ⁿ as n→∞

He couldn't calculate it exactly, but showed it had to fall between 2 and 3.

That limit is e ≈ 2.71828...

He didn't name it. Fifty years later, Leonhard Euler calculated it to 18 decimal places, and the letter e stuck - appearing first in Euler's letter to Goldbach in 1731.

Why e is special:

e is the only number where the exponential function eˣ is its own derivative.

That one property sounds abstract. In practice it means: when you differentiate eˣ, you get eˣ back. The rate of change equals the value itself.

This makes the calculus clean. No extra constants, no messy chain rule artifacts.

And clean calculus is everything when you're training a neural network through millions of gradient updates.

Where e lives in every AI system you build:

1/ Sigmoid function
the S-shaped function that maps any real number to a value between 0 and 1.

σ(x) = 1 / (1 + e⁻ˣ)

Used in logistic regression and binary classification. e is what makes its derivative expressible in terms of itself - σ'(x) = σ(x) · (1 - σ(x)), which keeps backpropagation computationally clean.

2/ Softmax
the function that converts raw model output scores (logits) into a probability distribution across multiple classes.

softmax(xᵢ) = eˣⁱ / Σeˣʲ

Every token probability your LLM produces passes through this. e is used because it keeps all values positive and makes the gradients during training well-behaved.

3/ Cross-entropy loss
the loss function used to train most classifiers and language models. It uses the natural logarithm (ln), which is the inverse of e.

The full picture:

A banker's question in 1683 → a universal mathematical constant → the backbone of every neural network trained today.

Bernoulli was thinking about interest rates. He had no idea he was laying the foundation for gradient descent.

That's what makes pure mathematics dangerous. It doesn't look useful until suddenly it's everywhere.

[steve hsu @hsu_steve]

Tim Gowers on a possible AI-dominated future of mathematics:

... because of [AI] people are no longer motivated to put in the years of effort it takes to reach the level of expertise that a typical research mathematician has now. After a decade or two, we might arrive at a situation where the mathematical literature has, in some form, been vastly expanded, but there is no corresponding community of human experts who have a shared understanding of parts of it.

https://t.co/ixBI0HNm0V

[Shreyas Doshi @shreyas]

If you are very smart and you operate as though execution vs. ideas are distinct phases, you might still get good at execution or good at ideas, but you will never get great at either.

Greatness requires a visceral understanding that execution and ideas are one thing, not two.

[Rohan Paul @rohanpaul_ai]

Boris Cherny, head of Claude Code at Anthropic, on optimizing token cost and model use.

“I use Fable for everything"

There is probably a 50% opportunity to reduce the investment (on token cost). However, there may be a 1,000%, 10,000%, or even 100,000% opportunity to increase the return.

Therefore, I would simply use the most expensive model and focus on how to get more out of it. Ask yourself, ‘How do I increase the return?’ Do not focus primarily on cost-cutting."

----

From "Scale" YouTube channel, (full video link in comment)

[Lance Martin @RLanceMartin]

lol this talk is from AIE 2024 (before I was at Anthropic). im glad graphs are cool (again)!

[Mahesh Sathiamoorthy @madiator]

Data curation (RL envs) is all you need for RSI

[Book Therapy @Book_therapy223]

Your environment always wins over your willpower, choose wisely.

[Y Combinator @ycombinator]

NVIDIA started with the wrong technology, learned the right one from three textbooks bought at Fry’s, and went on to invent most of the major breakthroughs in modern computing.

At Startup School 2026, YC's @garrytan sits down with @nvidia Founder and CEO @JensenHuang to talk about confronting reality, learning your way into new domains, and why resilience — getting through one day at a time — matters more than anything else.

01:07 — NVIDIA's wrong algorithm
03:14 — Buying textbooks to save the company
05:29 — The real big idea
07:06 — The Sega story
09:37 — The $300M IPO
10:33 — Seeing AlexNet differently
12:44 — Reinventing the full stack
13:30 — How to build a first-principles org
17:01 — Build the car to fit your driving style
19:12 — Frontier algorithms
20:50 — Systems thinking is the new coding
23:26 — Should you own your own AI?
25:43 — How NVIDIA sses agents internally
30:51 — AI and job creation
34:21 — The ChatGPT moment for robots
37:24 — Where physical AI shows up first
39:18 — Why Jensen just joined X
41:01 — What to learn that still matters
44:32 — The mindset you should have - "How hard can it be?"