Anastasios Skarlatidis

1/4 LLMs solve research grade math problems but struggle with basic calculations. We bridge this gap by turning them to computers. We built a computer INSIDE a transformer that can run programs for millions of steps in seconds solving even the hardest Sudokus with 100% accuracy

@weakwire mosh + tmux

2026 is the year to start using "mosh". Also give happy.engineering a try

Time is almost up! ⏰ Today is your last chance to make a gift that goes twice as far. Every donation before midnight helps us reach our $150,000 goal and fuels the research driving treatments and cures for blinding diseases. Don’t miss this opportunity to make a lasting impact before the year ends: bit.ly/3MsUG9m Video Description: Dark blue background with Foundation logo in the top left corner. Copy on the right reads, "END-OF-YEAR MATCH."

Nikolas Pontikos et al. @npontikos @UCLeye @MoorfieldsBRC study Current Applications of Artificial Intelligence for Fuchs Endothelial Corneal Dystrophy: A Systematic Review, tvst.arvojournals.org/article.aspx?a….

An open-source AI tool could improve the diagnosis timeline for #InheritedRetinalDiseases.

Read more about the recent report in @TheLancet of an effective treatment for the most severe form of childhood blindness, giving infants life-changing improvements in sight, in an interview with Prof @MichelMichaelid by @medresearchnews: medicalresearch.com/gene-therapy-r…

Transformers don't know how to reason. MLNs do: homes.cs.washington.edu/~pedrod/papers…

Steve jobs took 18 seconds to think about his answer to this great question

Thanks for the raspberry pi pico gift

Had a fantastic time visiting the @mouseio ! 🖥️💾 It was a great chance to share some tech nostalgia with my kids, showing them vintage computers and game consoles from before my time and my own childhood. So many memories and tech history under one roof #HellenicITMuseum

Just never give up.

New study evaluated a deep learning algorithm that automatically quantified multiple OCT features at multiple time-points in a large real-world dataset of patients treated for neovascular #AMD. @pearsekeane @g_nmoraes ow.ly/azwV50SAjcY

@dtsomp Υπομονή :/

"A survey on the evolution of stream processing systems" Banger of a paper by @MarioFragkoulis, discussing ordering semantics, fault tolerance, scalability, state management and more, of different stream processors. 👉 link.springer.com/article/10.100…

CRISPR-GPT uses LLMs to automate and enhance the design process of complex biological gene-editing experiments. Augments LLMs with domain knowledge, external tools and a task execution system. Uses CoT reasoning model and state machines. Ref: arxiv.org/pdf/2404.18021…

Sadly, every day I am writing less and less Scala code.

I’m desperately trying to write code and do machine learning, instead of using APIs everywhere. What is happening?

Few people and companies understand this: #Scala is that rare tool that makes everything easier: - backend - frontend (more than most JS libs, and I've used many) - architecture - domain modeling - testing - code reasoning (underappreciated and directly related to productivity)

Announcing the alpha release of torchtune! torchtune is a PyTorch-native library for fine-tuning LLMs. It combines hackable memory-efficient fine-tuning recipes with integrations into your favorite tools. Get started fine-tuning today! Details: hubs.la/Q02t214F0