Fouad Mardini

@LambrosPetrou I think you are referring to reinforcement learning with verifiable rewards arxiv.org/abs/2506.14245

Shower idea: What if the best models will be generated by blindly optimizing the weights based on benchmarks? The effort then is making good benchmarks (value functions?). Any model provider that does this already, maybe in secret?

@keenanisalive Richard Borcherds’ lectures on his YouTube channel are simply the best I’ve seen. Old school tech as well

What are some good examples of how to present detailed calculations / proofs that work *better* than a chalkboard? I often hear things like "I can't *really* follow/understand it unless it's on a chalkboard." But that's not helpful online—and limits access to education.

We're very proud at @MaqsamHQ to present our #Arabic Speech to Text, that is beating Google and Microsoft when it comes to spoken Arabic. If you speak Arabic, give it a go: intelligence.maqsam.com

I often get asked "What should I do to #grow in my career?" I recently realized that #seniority in different functions have a lot in common. Abstractly, I identify 5 levels of seniority that I illustrate below (with a simple scenario of a leader wanting to cook a burger): 1/7

@sinantaifour @garybernhardt In the same vein github.com/codecrafters-i…

Just came across the "From Scratch" screencasts from @garybernhardt, where he walks you through implementing toy versions of everyday programming tools (e.g. compiler, shell) in 30 minutes. destroyallsoftware.com/screencasts/ca…

🔍 Question for AWS users 👇 Do you use any tools that simplify and make it easy to manage your applications, including AWS resources, deployments, host setup, etc? Something like Forge for Laravel, but dedicated to AWS, and allowing generic AWS applications. #aws #automation

@sinantaifour Fargate, no?

I want a service that has one API, where I push a #Docker image, and it runs it to termination.

@DanielJ20335910 @fchollet I think this is the key question here.

@fchollet Is there any DL models that are "not" interpolative?

The biggest difficulty people (even fairly senior folks) seem to have in grasping that most deep learning models perform interpolation is that they think "interpolation" means "input-space L2 interpolation", i.e. linear regression 🤦

Just watched the "Observability for emerging infra" talk by @mipsytipsy. Even though I am a big proponent for testing before production, I do agree with all the points raised! youtube.com/watch?v=fOdtgH… My favourite points raised 👇

@fdahdaleh Congratulations Fadi 🇺🇸🇺🇸

I arrived at the United States around 13 years ago and, today, I became a US citizen. I am extremely grateful for the opportunities and life this country has given me and my family.

@EscardoMartin What are your favourite books on the subject?

@5p4k I guess if you don't really care about performance, you can use blocking calls. You can also structure your code to have the async parts, call into "pure" functions that simply do compute and have these be async/await free.

@5p4k This is due to the JavaScript execution model where it never blocks. developer.mozilla.org/en-US/docs/Web…

@sinantaifour @jawaker @hajhasan Thank you Sinan! I have no doubts that @MaqsamHQ will grow and mature into a bigger success, after all it's run by the best. Looking forward to a day where we work again together

Huge congrats to @jawaker for getting acquired! I was the first employee in @jawaker, @hajhasan and @fmardini were clearly world class. I learned so much. It's where I caught the startup bug, and not for @jawaker I probably wouldn't have cofounded any startup.

@sinantaifour I'm sure they mean relativistic mass as you approach the speed of light

The official driver's handbook in #Jordan claims in a section titled Earth Gravity: "according to the law of kinetic energy the faster a vehicle the lighter it is". Gotta love how scientific they are! I wish this was the biggest nonsense in this book.

@sinantaifour @MaqsamHQ Congratulations man! Great work

Back in 2019, we created a financial model as we were raising @MaqsamHQ's F&F round. We ended up outperforming the optimistic scenario of the model, which at the time sounded very optimistic. Our updated model now foresees us at a high-8-figure annual revenue by 2023. Exciting!

@JSEllenberg @mcopelov (from) the page of the journalist Nawaf Nabari

@mcopelov What’s the caption?

@gabrielpeyre Such a nice result, thanks for sharing. I like how straight forward the proof is as well.

Oldies but goldies: S. Gerschgorin, Über die Abgrenzung der Eigenwerte einer Matrix, 1931. Gershgorin circle theorem locates the eigenvalues of a matrix in the union of disks centered at the diagonal entries. en.wikipedia.org/wiki/Gershgori…

Thanks for sharing, I greatly enjoyed the lecture by Paul R Halmos.

@sinantaifour @smdiehl Not really, e.g. there is no scenario where cryptos will ever pay dividends. Stocks over long time periods have values reflecting their worth (Buffett's "weighing machine"). What's the intrinsic worth of cryptos?

@smdiehl Wouldn't the same logic apply to stocks too? Except for IPOs, when you're buying a stock it's because someone else is selling it.

The Bitcoin Free Lunch Problem: People think they're investing (on paper) but they're really just handing their dollars off to people who are running away with it, theres no economic way that system can generate returns. It's just redistribution from fools to greater fools.