Giulio Sighieri

tech / ml | sensitive young man
MSc Data science - @unibocconi
Ambassador @italianbldrs
AI engineer for

[Giulio Sighieri @techy_windy]

THIS IS THE FUTURE I'M SO IN FOR THIS SHIT LET'S GO YOU DID A GREAT JOB GUYS I'M WORKING HARD AS FUCK I WANT TO WORK ON THESE STUFF TOO

🚀🚀🚀🚀🚀🚀🚀🚀🚀

[Odyssey @odysseyml]

Today we're launching Odyssey-3, the most powerful foundation world model yet.

It sets a new state of the art on Physics-IQ, and powers robots, trains AIs, and generates interactive experiences. It's really cool.

Experience the model today, all for free!

[Giulio Sighieri @techy_windy]

Since they improved the speed of @bot i cannot honestly stop using it

is wonderful!

[Giulio Sighieri @techy_windy]

Very good point, recommended reading

Personally, I love game theory problems.
I think I'd like a lot to train game theory models

[sunil pai @threepointone]

The thing I learned this year is that when engineers say they like working on “hard problems”, they actually don't. What they like is puzzle solving, and having a repertoire of patterns that they can pattern match onto traditionally hard problems, usually involving databases or something with scale, etc. And they know that the industry used to pay you a lot for how deep and broad this repertoire was. Unfortunately, in 2026, well, you have a cheat code machine that has infinite breadth and depth for pattern matching. So the question is, do you actually like hard problems? Because there are now hard problems that are a complection of engineering, product, society, culture, politics, and they are fairly intractable, probably unsolvable! Because if you do want to work on those actually hard problems, there is so much money that people will pay you for it. But you’ll have to face that truth about yourself first.

[Giulio Sighieri @techy_windy]

I believe they are secretly in love with each other btw

[Giulio Sighieri @techy_windy]

This is crazy wtf

[Domenico 🐺 @dom_gag_96]

🚨Oggi lanciamo SuperPino.

SuperPino non si limita a rispondere alle domande.

Inizia a fare cose per te.

Gli mandi una foto di una buca, di un lampione spento o di un problema nel tuo comune.

Lui capisce dove sei, individua l’amministrazione competente, prepara la segnalazione e te la fa approvare.

Poi la invia e ti permette di seguirne l’andamento.

Ma questa è solo la prima cosa che sa fare.

L’idea dietro SuperPino è molto più grande: diventare il punto di contatto tra te e i servizi pubblici.

Tu dici cosa vuoi fare:

“Ho cambiato casa.”
“Voglio aprire un’attività.”
“Ho diritto a qualche agevolazione?”

SuperPino capisce cosa serve, trova il servizio giusto e ti aiuta a portare avanti il processo.

Oggi una segnalazione. Domani, sempre più servizi pubblici.

Prima abbiamo reso i dati più accessibili.

Poi le informazioni.

Ora iniziamo a lavorare sulle azioni.

Questo è SuperPino. Ed è solo l’inizio.

Provalo su https://t.co/uJAhgJ5oLS

[Giulio Sighieri @techy_windy]

@Full_Metal_QR I've been in that stuff for almost two years.

Having a good friend that really cared about me is what saved my life. Not joking at all

Sun in the darkness

[Giulio Sighieri @techy_windy]

4D chess only on this app btw

[Giulio Sighieri @techy_windy]

@calebfahlgren Check out my /study-book @calebfahlgren, has been inspired by your work and I think you might like it a lot!

https://t.co/UMtcLa0L6E

[Giulio Sighieri @techy_windy]

/study-book

I had claude write me a 519-page textbook on neural networks. Not a summary: a whole book covering at good depth all the fundamental knowledge in NNs.

It's an open-source skill. Inspired by @calebfahlgren's recent post, I wanted to learn anything with material made exactly the way I want it.

You tell it what you want to learn, pick a depth from 0 to 5, and optionally give it your own sources: slides, notes, etc... It plans the book, you approve the plan, then:

→ one writer agent per chapter
→ auditor fact-checks every chapter
→ "student" agent reads it and flags where it got lost
→ a final review of the whole book

Then a typesetting engine I built turns it into a real book: justified text, figures, cross-references, glossary, index.

The result: 403 of 403 topics covered, every number computed by a script, every quote checked against its source.

By being a passionate learner, I'm more excited than ever before!

You can find the skill and engine in the comments!

Video by Opus 5.5

PS: is quite token hungry!

[Giulio Sighieri @techy_windy]

@j4orz you might want to take a look at this if interested then

https://t.co/UMtcLa1iWc

[Giulio Sighieri @techy_windy]

/study-book

I had claude write me a 519-page textbook on neural networks. Not a summary: a whole book covering at good depth all the fundamental knowledge in NNs.

It's an open-source skill. Inspired by @calebfahlgren's recent post, I wanted to learn anything with material made exactly the way I want it.

You tell it what you want to learn, pick a depth from 0 to 5, and optionally give it your own sources: slides, notes, etc... It plans the book, you approve the plan, then:

→ one writer agent per chapter
→ auditor fact-checks every chapter
→ "student" agent reads it and flags where it got lost
→ a final review of the whole book

Then a typesetting engine I built turns it into a real book: justified text, figures, cross-references, glossary, index.

The result: 403 of 403 topics covered, every number computed by a script, every quote checked against its source.

By being a passionate learner, I'm more excited than ever before!

You can find the skill and engine in the comments!

Video by Opus 5.5

PS: is quite token hungry!

[Giulio Sighieri @techy_windy]

The future is bright

[Neuralink @neuralink]

Our clinical trial participants have now used their Neuralink implants for over 50,000 hours, building one of the largest datasets of intracortical neural activity ever.

We used it to pretrain a foundation model for BCI: self-supervised neural encoders that turn noisy, drifting spikes into stable representations.

Some initial results:
• Decoders that keep working for weeks instead of days. For some users, calibration dropped from 10 min/day to 10 min/week
• Teleporting cursors that jump straight to the intended target
• A new cursor-control record of 11.32 bits per second (exceeding the performance of most able-bodied people)

We’re just getting started. Future directions include pooling data across participants, one-shot (or zero-shot) calibration, and an API for the motor cortex.

Read more:
https://t.co/sIQXFUI7pc

[Giulio Sighieri @techy_windy]

@Full_Metal_QR Time to chinamaxx

[Giulio Sighieri @techy_windy]

Usage on Opus 5.5 has been massively reduced.
Last week I could start 10 agents for hours and consume 20% of the 5 hours in 1 hour, now 2 agents + 10 minutes...

something's off

[Giulio Sighieri @techy_windy]

Best 2026 video

[Charles Curran @charliebcurran]

I used AI to explain the AI Doomer drama, with Lord of the Rings and kittens.

[Giulio Sighieri @techy_windy]

What I noticed is that often the best engineers working for the best companies post 5 times or less a year.

They are probably busy working instead of yapping.

Lesson in there

[Giulio Sighieri @techy_windy]

You can find skill + engine + NN book example at

https://t.co/kmoyrWPWVR

:D

[Giulio Sighieri @techy_windy]

/study-book

I had claude write me a 519-page textbook on neural networks. Not a summary: a whole book covering at good depth all the fundamental knowledge in NNs.

It's an open-source skill. Inspired by @calebfahlgren's recent post, I wanted to learn anything with material made exactly the way I want it.

You tell it what you want to learn, pick a depth from 0 to 5, and optionally give it your own sources: slides, notes, etc... It plans the book, you approve the plan, then:

→ one writer agent per chapter
→ auditor fact-checks every chapter
→ "student" agent reads it and flags where it got lost
→ a final review of the whole book

Then a typesetting engine I built turns it into a real book: justified text, figures, cross-references, glossary, index.

The result: 403 of 403 topics covered, every number computed by a script, every quote checked against its source.

By being a passionate learner, I'm more excited than ever before!

You can find the skill and engine in the comments!

Video by Opus 5.5

PS: is quite token hungry!

[Caleb @calebfahlgren]

My favorite way to learn with AI right now: ask for a dense technical manual on the topic, grounded in the actual spec and source code with diagrams.

Made these for Parquet and GGUF. Frontier models are shockingly good at this now.

[Giulio Sighieri @techy_windy]

Crazy speed + crazy work

These guys are unstoppable, let's fucking go 🇮🇹

[Domenico 🐺 @dom_gag_96]

Oggi lanciamo https://t.co/H7Aau08wY4

L’idea è di poter affidare a un agente AI le pratiche con la pubblica amministrazione, direttamente da una chat.

«Ho cambiato casa.»

«Voglio aprire un’attività.»

«A quali agevolazioni ho diritto?»

L’obiettivo è arrivare a un agente che capisce la richiesta, individua i servizi, prepara le pratiche e, con la tua autorizzazione, le porta avanti.

Tutto a partire da una chat.

Dopo aver visto https://t.co/rE7bIAIYvi, abbiamo deciso di provarci anche per l’Italia, partendo dalla base che abbiamo costruito nell’ultimo mese lavorando a DoveVannoINostriSoldi.

ItaliaAperta è un progetto indipendente.

Non abbiamo potere decisionale su come funziona la PA, ma speriamo che mettere online qualcosa da provare possa spingere le istituzioni a prendere sul serio questa possibilità.

Accedere ai servizi pubblici dovrebbe essere semplice quanto scrivere in chat.

Vorremmo che diventasse normale anche in Italia.

È online la primissima versione.

C’è ancora molto da fare: provatela, diteci cosa manca e condividetela.→ https://t.co/CAlVCX8LB7

[Giulio Sighieri @techy_windy]

@leothecurious @paradigmainc @tensorqt @alxfazio Glad you liked it :D

[Giulio Sighieri @techy_windy]

I liked a lot the launch of Violetto 1B by @paradigmainc, plus I hated how they got zero media coverage.

I had some tokens left, so I said "Opus, time to make a banger video for a banger model"

50 minutes later:

@tensorqt @alxfazio hope you like it

[Giulio Sighieri @techy_windy]

@tensorqt I pick the eternal return in the middle thanks

[Giulio Sighieri @techy_windy]

@Full_Metal_QR Banger quote btw