AI & Tech Developments - Feb 28

  • 04:32 β€” We’re excited to introduce Doc-to-LoRA and Text-to-LoRA, two related research exploring how to make LLM customization faster and more accessible. @SakanaAILabs
  • 12:01 β€” Pewdiepie just trained his own LLM. His model outperformed deepseek v2.5, LLAMA-4 and GPT-4o in coding benchmark. @birdabo
  • 12:46 β€” Qwen3.5 shows drastic performance improvements with fewer resources. @sudoingX
  • 15:10 β€” I took 1 hour to automate 80% of our meta ads audit reporting for prospective clients. @peterczepiga
  • 15:45 β€” Holy shit… Someone just built the ultimate prompt collection for AI image creators. @hasantoxr
  • 16:10 β€” This is how to make your AI 10x more useful: Give your agent the ability to parse the whole Internet. @svpino
  • 20:30 β€” i am pretty sure codex /review and claude review surpasses tools like greptile and coderabbit. can people confirm? @dejavucoder
  • 20:37 β€” json-render can now render images. @ctatedev
  • 20:53 β€” It’s cool how Hegseth is starting a two-front war on both human and machine intelligence. @luke_metro
  • 23:40 β€” We seem close to: Give an agent access to a competitor app on a computer… @Suhail

πŸ“± Source Tweets

Holy shit...A single GitHub repo just made Claude Code 10x more powerful.

β€” @aiwithjainam

OpenClaw just launched Mission Control. This changes everything.

β€” @JulianGoldieSEO

5 days ago it took 2 GPUs to build this. today it takes 1.

β€” @sudoingX

as of this writing claude is now #2 in the app store & there’s a real non trivial downside scenario here for openai that many aren’t really grasping.

β€” @signulll

Cool chart showing the ratio of Tab complete requests to Agent requests in Cursor.

β€” @karpathy

It's too bad Anthropic is completely EA/autist. They don't have a normie who can interface with the administration...

β€” @brianluidog

We’re excited to introduce Doc-to-LoRA and Text-to-LoRA, two related research exploring how to make LLM customization faster and more accessible.

β€” @SakanaAILabs

Pewdiepie just trained his own LLM. His model outperformed deepseek v2.5, LLAMA-4 and GPT-4o in coding benchmark.

β€” @birdabo

Qwen3.5 shows drastic performance improvements with fewer resources.

β€” @sudoingX

I took 1 hour to automate 80% of our meta ads audit reporting for prospective clients.

β€” @peterczepiga

Holy shit... Someone just built the ultimate prompt collection for AI image creators.

β€” @hasantoxr

This is how to make your AI 10x more useful: Give your agent the ability to parse the whole Internet.

β€” @svpino

i am pretty sure codex /review and claude review surpasses tools like greptile and coderabbit. can people confirm?

β€” @dejavucoder

json-render can now render images.

β€” @ctatedev

It’s cool how Hegseth is starting a two-front war on both human and machine intelligence.

β€” @luke_metro

We seem close to: Give an agent access to a competitor app on a computer...

β€” @Suhail