Julien Salinas (@juliensalinasen) 's Twitter Profile
Julien Salinas

@juliensalinasen

CTO at NLP Cloud. Full-Stack engineer and AI specialist. Boxing instructor.
Reach our for expertise on AI implementation and full-stack engineering ⚡️

ID: 940503161760075776

linkhttps://nlpcloud.com calendar_today12-12-2017 08:46:42

298 Tweet

299 Takipçi

191 Takip Edilen

Julien Salinas (@juliensalinasen) 's Twitter Profile Photo

Claude Code has recently emerged as a powerful alternative to Cursor and Windsurf. Launched in February 2025, Claude Code is a terminal-based AI coding assistant powered by Claude. Unlike traditional IDE-centric tools, developers interact with Claude Code in a terminal. Claude

Mario Nawfal (@marionawfal) 's Twitter Profile Photo

WE USED TO SLEEP TWICE A NIGHT - UNTIL THE INDUSTRIAL REVOLUTION RUINED IT For thousands of years, humans had two sleeps: one at sundown, a wakeful “watch” around midnight, then another stretch till dawn. That midnight wake-up? Totally normal. People prayed, read, made beer,

DHH (@dhh) 's Twitter Profile Photo

Can't believe Elon actually did it. FSD has gotten shockingly, eerily good. Just did another drive from Malibu to LAX without taking the wheel once. Plenty of tricky, subtle situations along the way. What a time to be alive!

Julien Salinas (@juliensalinasen) 's Twitter Profile Photo

Cloudflare is at the right place to help filter AI crawling. Implementing a system that asks AI crawlers to pay for the data they harvest seems to be a fair decision in my opinion. Now let's see how they implement this idea...

Cloudflare is at the right place to help filter AI crawling.

Implementing a system that asks AI crawlers to pay for the data they harvest seems to be a fair decision in my opinion.

Now let's see how they implement this idea...
@levelsio (@levelsio) 's Twitter Profile Photo

SpaceX Earth to Earth passenger flights are now pending approval 27,000km/h Hong Kong to Singapore: 22 min LA to Toronto: 24 min Bangkok to Dubai: 27 min Tokyo to Singapore: 28min London to Dubai: 29min London to New York: 29min New York to Paris: 30min Honolulu to Tokyo: 30

Avi Chawla (@_avichawla) 's Twitter Profile Photo

A new embedding model cuts vector DB costs by ~200x. It also outperforms OpenAI and Cohere models. Here's a complete breakdown (with visuals):

DHH (@dhh) 's Twitter Profile Photo

Codie Sanchez Probably true when internal drive is lacking or the culture is built on supervision. But it's possible hire folks with high internal drive and build a culture on outcomes. Then work-from-home does wonders for the right people.

DHH (@dhh) 's Twitter Profile Photo

Omarchy 2.0 now installs in under 5 minutes on a fresh machine with a fast connection: 1 min: Formatting + Encrypting drive 1 min: Installing Arch itself 2:40min: Installing + configuring all of Omarchy FIVE MINUTES AND YOUR COMPUTER IS READY TO ROCK! 🤘

Andrej Karpathy (@karpathy) 's Twitter Profile Photo

In era of pretraining, what mattered was internet text. You'd primarily want a large, diverse, high quality collection of internet documents to learn from. In era of supervised finetuning, it was conversations. Contract workers are hired to create answers for questions, a bit

Rohan Paul (@rohanpaul_ai) 's Twitter Profile Photo

BRILLIANT Google DeepMind research. Even the best embeddings cannot represent all possible query-document combinations, which means some answers are mathematically impossible to recover. Reveals a sharp truth, embedding models can only capture so many pairings, and beyond that,

BRILLIANT <a href="/GoogleDeepMind/">Google DeepMind</a> research.

Even the best embeddings cannot represent all possible query-document combinations, which means some answers are mathematically impossible to recover.

Reveals a sharp truth, embedding models can only capture so many pairings, and beyond that,
Aleksa Gordić (水平问题) (@gordic_aleksa) 's Twitter Profile Photo

New in-depth blog post - "Inside vLLM: Anatomy of a High-Throughput LLM Inference System". Probably the most in depth explanation of how LLM inference engines and vLLM in particular work! Took me a while to get this level of understanding of the codebase and then to write up

New in-depth blog post - "Inside vLLM: Anatomy of a High-Throughput LLM Inference System". Probably the most in depth explanation of how LLM inference engines and vLLM in particular work!

Took me a while to get this level of understanding of the codebase and then to write up
Eric Larchevêque (@ericlarch) 's Twitter Profile Photo

Les français plébiscitent la taxe Zucman à 86%. C’est un véritable raz-de-marée transpartisan. Il n’existe plus de droite libérale en France, et surtout il ne semble plus y avoir de français intéressés par un autre discours que celui de taxer les plus riches. La gauche a réussi

Les français plébiscitent la taxe Zucman à 86%.
C’est un véritable raz-de-marée transpartisan.

Il n’existe plus de droite libérale en France, et surtout il ne semble plus y avoir de français intéressés par un autre discours que celui de taxer les plus riches.

La gauche a réussi
Pratham (@prathkum) 's Twitter Profile Photo

Coding was never really about writing code. It was about breaking a problem into steps, imagining the edge cases, and knowing where trade-offs live. AI just took away the boring part. What’s left is the real work: thinking clearly.

Julien Salinas (@juliensalinasen) 's Twitter Profile Photo

After a couple of years on Mac, I'm finally back to Linux thanks to DHH ! Omarchy's interface is very nice and fast, and keyboard navigation is very efficient. It's both a pleasure and a productivity booster. Before my Mac experience I had spent 15 years on Ubuntu, Fedora,

After a couple of years on Mac, I'm finally back to Linux thanks to <a href="/dhh/">DHH</a> !

Omarchy's interface is very nice and fast, and keyboard navigation is very efficient. 
It's both a pleasure and a productivity booster.

Before my Mac experience I had spent 15 years on Ubuntu, Fedora,