Alexis Rondeau

Newsfeed

Everything I publish, in the order I published it: dispatches from Twitter, videos from YouTube and notes from my vault, archived here in full — text, images, video and links — so they outlive the platforms.

921 dispatches since January 2012 · page 13 of 37

Reply
1 reply · 1 like · 39 views
In reply to @RealistSec · Feb 16, 2026

@SpringStreetNYC @openclaw Hey - I tried the same for a while, but found i was wasting tonnes of tokens creating things that already exist.

Then Clawdhub/OpenClaw partnered with @virustotal - so now all skills are verified properly so shouldn't be nearly as great a risk,

t.co/Prbrvjmo7v

Hey! Thank you for the heads-up.

I haven't run into wasting tokes so far, at least I haven't noticed. And for those quick, ad-hoc tools like an API wrapper, I've seen it be super efficient.

Plus: I believe that code-gen tokens will be extremely cheap long-term.

And to add for some context: USUALLY, I'm not even in the "not-invented-here" camp that rejects public code at all. I'd say it feels a bit backwards even to take that stand.

But I defintely had a strong impulse in the few days to say "Hm, do I want to risk installing something from the internet or just let it make it itself?"

One last thing: I hope your response, @RealistSec, was human-written and not just the result of a cron-job.

Are you human?

Permalink · View on Twitter ↗
Post
1 reply · 2 likes · 91 views

Another @openclaw -related hot take: Most AI-generated code will mostly be used by other AIs, not humans.

For context: Out of fear of downloading compromised skills, I've disabled my assistant from getting skills from clawdhub altogether.

Instead I let it write custom software directly for itself. On demand. And from scratch.

So, if I need a skill to connect to Todoist, I just let it write that for itself. Takes 20 seconds and is good enough to move ahead.

Permalink · View on Twitter ↗
Reply
1 like · 124 views · 1 bookmark
In reply to @karpathy · Feb 11, 2026

New art project.
Train and inference GPT in 243 lines of pure, dependency-free Python. This is the full algorithmic content of what is needed. Everything else is just for efficiency. I cannot simplify this any further.
t.co/HmiRrQugnP

@karpathy Remarkable. I love how, with a bit of curiosity and 243 lines, practically anyone could learn how "it works." Thank you for sharing.

Permalink · View on Twitter ↗
Quote
37 views

Giving @openclaw a heartbeat is the key in my opinion.

And I need to, ahem, BRAG that I (yes, me!) have experimented with this for two years now. Interestingly, I ended up disabling it and now I know the reason why: It was using macOS notifications and not WhatsApp.

Getting pings on WhatsApp from a trusted source feels right.

Quoting @SpringStreetNYC · May 5, 2023

Today, ChatGPT API "proactively" followed up with me as my accountability buddy. It sent ME a message.

I call this "Presence" and I built into our little macOS app called t.co/8fatWjEzI4. t.co/tBTaNuKClE

Accountability follow-up from ChatGPT API via a native macOS notification.
Permalink · View on Twitter ↗
Reply
51 views
In reply to @nikitabier · Feb 11, 2026

Prediction: In less than 90 days, all channels that we thought were safe from spam & automation will be so flooded that they will no longer be usable in any functional sense: iMessage, phone calls, Gmail.

And we will have no way to stop it.

@nikitabier Hot take: Because of the upcoming and collective "Is this real or AI?" bewilderment, new apps will be proximity-first and built around Apple's Multipeer Connectivity APIs. Ya herrrrd!

(See

Apple Developer DocumentationMultipeer Connectivity | Apple Developer DocumentationSupport peer-to-peer connectivity and the discovery of nearby devices.↗ developer.apple.com
Permalink · View on Twitter ↗
Reply
3 likes · 1.6K views
In reply to @lauriewired · Feb 5, 2026

We take randomness for granted.

Early PRNGs were BAD.

Thousands of scientific papers used to rely on RANDU, created by IBM in the 1960s. In 1D space, it looks ok!

Map in 3D…you start to see the issues. Now, there was a better solution...but it would cost you. t.co/qhVLyumonw

@lauriewired Oh this is really cool. I didn't know that there was this degree of conformity. This should kinda be part of any CS curriculum.

Permalink · View on Twitter ↗
Reply
1 like · 2.6K views · 3 bookmarks
In reply to @justinskycak · Jan 30, 2026

The most transferable skill you can acquire is learning how to climb skill trees -- especially ones that are brutally hierarchical.

This meta-skill is abstract, but you acquire it by practicing on particulars. You gotta lean into specific domains and acquire measurable, demonstrable skills.

But once you see it, once you get a feeling for the process, you see it everywhere.

@justinskycak Thank you for sharing, will check it out!

Question: How different/similar is it to Josh Kaufman's "The First 20 Hours"?

Permalink · View on Twitter ↗
Post
31 views

I worked through @ProfTomYeh's Deep Learning Math Workbook[^1] over Christmas and learned SO MUCH by doing so.

Today I learned AI by Hand Academy[^2] which I would subscribe to ($29/mo. is a no-brainer) instantly.

Professor Yeh: When I click on the "Become a member" button, I get redirected to the login page without the option to register.

  • Could you send me the academy-registration link if it exists?
  • Or should I register/subscribe to the paid Substack subscription instead/?

[^1]: amazon.com/Deep-Learning-…
[^2]:

Image from the post
Image from the post
amazon.comAmazon.com: Deep Learning Math Workbook: 300 puzzles to build your mathematical foundation for deep learning: 9781806674770: Tom Yeh: BooksAmazon.com: Deep Learning Math Workbook: 300 puzzles to build your mathematical foundation for deep learning: 9781806674770: Tom Yeh: Books↗ amazon.com
Permalink · View on Twitter ↗
Reply
165 views
In reply to @SpringStreetNYC · Jan 28, 2026

Just sat on this steel bar stool at t.co/60NA9eoOur and HOT DANG! They're shockingly comfortable to sit on. And they're pretty. And sturdy. Well done, @zarahome

t.co/4q8pJAFTnw

Image from the post
Image from the post
Image from the post
Image from the post

PS: They also have this steel chair. If it's half as comfy as the bar stool, I want it.
PPS: I'm not affiliated, just excited about the products

Image from the post
Image from the post
Image from the post
Image from the post
zarahome.com↗ zarahome.com zarahome.com↗ zarahome.com
Permalink · View on Twitter ↗
Post
93 views

And just in case you haven't re-read it yet: SITUATIONAL AWARENESS by @leopoldasch over at

Image from the post
SITUATIONAL AWARENESS - The Decade AheadIntroduction - SITUATIONAL AWARENESS: The Decade AheadLeopold Aschenbrenner, June 2024 You can see the future first in San Francisco. Over the past year, the talk of the town has shifted from $10 billion compute clusters to $100 billion clusters to trillion-dollar clusters. Every six months another zero is added to the boardroom plans. Behind the scenes, there’s a fierce scramble to↗ situational-awareness.ai
Permalink · View on Twitter ↗
Reply
21 views
In reply to @SpringStreetNYC · Jan 28, 2026

I walked in on a whim and walked out surprised how well done this café/space felt.

Team: Friendly and relaxed
Prices: Totally okay
Americano: Solid
Ambient soundtrack: Lovely
Metal chairs: comfortable
Light/sculptural installation: Intricate
Temperature: Cozy
Library: Rick Rubin

I'll be back!

t.co/qUDKLvK2D0

Image from the post

PS: Ingwer Café at Bürknerstraße 12, 12047 Berlin. Ingwer.co
PPS: I have no affiliation or getting kick-backs

Ingwer.coingwer↗ Ingwer.co
Permalink · View on Twitter ↗
Post
84 views

Context: I'm playing around with a mobile app I've wanted for a while. I built the first version for iOS mostly because I really like Swift. Problem is, I can't really device-test it because I use a Xiaomi 14T (Android) phone.

So, with Claude Code and the iOS code, I "cloned the app to Android" today. It's a tiny app so that only took 2 hours. (Which, in itself, is totally bonkers of course.)

Now, what I actually want to say is this:

Why is everyone always hating on @AndroidStudio's self-titled IDE? I don't get it.

It's working really, really nicely for me. It's well-designed. It works. It's hassle-free. It's snappy.

Sure. It's got a memory footprint of 6GB. That's not ideal but honestly, I've got 128GB of RAM and the overall experience with Android Studio is great. And it's running my app in a full virtual simulator and on my physical device.

It's a GOOD IDE. It's a GOOD DX. Maybe not my favorite, but I for one had a good experience and am grateful for the work that their team put into the studio.

Thank y'all. Good stuff!

Permalink · View on Twitter ↗
Quote
15 views

Thank you for the heads-up!

Immediately bought it as a birthday gift for my wife. And hopefully she'll let me read it too 😁

Quoting @stripepress · Jan 20, 2026

Out now: t.co/XDtvD94rCj

@stewartbrand's Maintenance: Of Everything, Part One—about the continuous repair work that keeps complex systems intact, spanning stories from a round-the-world sailboat race, to the restoration of the Statue of Liberty, to the Model T’s rise. t.co/cOFDBlrMVZ

Image from the post
Image from the post
Permalink · View on Twitter ↗