←
Recursive18 posts
Recursive avatar

Recursive

@Recursive_SI

Recursive self-improving superintelligence to automate knowledge discovery.

◷ Joined May 2026
Recursive avatar
Recursive@Recursive_SI·Oct 21

Recursive Flint is here: our first model, built with the Eureka Machine. Fast, local coding. Open weights available on Hugging Face.

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

2:41
View quoted post ↗
Richard Socher avatar
⟳ Recursive retweeted
Richard Socher@RichardSocher·Oct 21

"I used Flint to build a bunch of programs and found it to be both fast and very capable. In my experience, models of this size can be annoyingly error-prone or slow compared to Claude or Codex, but Flint nailed my coding requests every time." — @willknight, @WIRED wired.com/story/recursive-se…

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

2:41
View quoted post ↗
Recursive avatar
Recursive@Recursive_SI·Oct 21

"I used Flint to build a bunch of programs and found it to be both fast and very capable. In my experience, models of this size can be annoyingly error-prone or slow compared to Claude or Codex, but Flint nailed my coding requests every time," writes @willknight.

WIRED@WIREDOct 21

Recursive makes the case that the benefits of using AI to build AI outweigh the risks. buff.ly/xsPUDpl

View quoted post ↗
NVIDIA RTX Spark avatar
⟳ Recursive retweeted
NVIDIA RTX Spark@NVIDIARTXSpark·Oct 21

Congrats @Recursive_SI on launching Flint 40B MoE trained on NVIDIA GB300 NVL72! 🙌 The open weight model was built with Eureka Machine, their automated AI research system that proposes ideas, tests them, and learns what to try next, all in under 4 months. Run the model locally on RTX Spark, DGX Spark, and RTX PRO 6000 Blackwell, with the NVFP4 checkpoint on @huggingface: huggingface.co/RecursiveSI/f…

Introducing Flint, @Recursive_SI’s first model: a fast local coding model built with our automated research system, the Eureka Machine. • Best-in-class fast local coding • Up to 2x faster than comparable models: 125+ tokens/sec on a MacBook Pro, 150+ on RTX Spark • Built from scratch in under 4 months Private, unmetered intelligence on your own machine. Open weights available now on @huggingface, run it locally on your favorite framework or try it for free on @opencode or @pidotdev.

2:41
View quoted post ↗
Josh Tobin avatar
⟳ Recursive retweeted
Josh Tobin@josh_tobin_·Aug 27

A fun small win for automated research: We found and helped fix some edge cases that could affect inference performance in vLLM and SGLang. In our last blog post, we described a reward hacking judge we developed for performance optimization tasks. While applying the judge to some new autoresearch work, it discovered that some FlashInfer (a library underpinning vLLM and SGLang) kernels used a hard coded value of -50,000 as a masked-attention sentinel -- even though valid QK values can be smaller. Corner cases like this that are numerically wrong but silent can cause a huge amount of headache to find and fix, e.g., the historical debate around flash attention (arxiv.org/abs/2405.02803). Great use case for AI. github.com/flashinfer-ai/fla…

Kanishka Narayan MP avatar
⟳ Recursive retweeted
Kanishka Narayan MP@KanishkaNarayan·May 15

It was such a privilege to join cofounders, @_rockt and Alexey Dosovitskiy, and a phenomenal team to celebrate @Recursive_SI’s launch and $650m raise last night. I have been struck by Tim’s exceptional ambition, experience and humility in the last few months. I am delighted he and team will be pushing the frontier of AI right here in the UK 🇬🇧

Post attachment