korrents
Salvatore Sanfilippo

What Salvatore Sanfilippo thinks about LLMs

@salvatore-sanfilippo · 20 positions · On ppll ↗

Programmer; wrote Redis and hping, and blogs about C, systems programming and working alone.

Salvatore Sanfilippo did not write this page.

We collected these quotes from things they published elsewhere, and every quote links to where it was said. They have no account here and have not endorsed this site. Quotes are word for word; the short line under each one is our own restatement, not their wording. Their own site. Is this you? Claim it or ask us to remove it. Or tell us what is wrong here.

8 dated positions, 2026, in their own words. Our reading of what Salvatore Sanfilippo has said — not written or endorsed by them.

  1. The real risk is leaks, not releases, and leaks happen inside frontier companies.

    ↗The real AI risk is inside the labsantirez.com 2nd of 3 in this piece

  2. 3 days earlier
  3. coding agentsdesign

    Automatic programming, instead, in the hands of people that are expert technicians, or expert programmers, expert designers, expert software architects, is to assume the role of Linus, with the agents and the LLMs assuming the role of the different maintainers of the different subsystems.

    ↗Being Linux Torvaldsantirez.com 3rd of 3 in this piece

  4. 12 days earlier
  5. LLMs are very good at writing locally optimal code, and are worse (but improving) with big ideas.

    ↗Control the ideas, not the codeantirez.com 3rd of 3 in this piece

  6. 5 weeks earlier
  7. code generation

    I have the feeling that the introduction of automatic QA may raise the bar of quality for new releases of software, and maybe partially compensate for the lower quality of the code produced at high speed with the use of automatic programming.

    ↗A new era for software testingantirez.com 3rd of 3 in this piece

  8. 13 days earlier
  9. Open weights models are now in a golden age, we have plenty and many are very powerful.

    ↗Distributing LLM inference in DwarfStarantirez.com 2nd of 2 in this piece

  10. 6 days earlier
  11. This means, basically, that just using line numbers is very fragile: to say, change line 22 with new="foobar" is not good.

    ↗Alternatives for the EDIT tool of LLM agentsantirez.com

  12. 5 days earlier
  13. It is clear that there was a need for single-model integration focused local AI experience

    ↗A few words on DS4antirez.com 1st of 3 in this piece

  14. The space will be occupied, in my vision, by the best current open weights model that is *practically fast* on a high end Mac or "GPU in a box" gear (like the DGX Spark and other similar setups).

    ↗A few words on DS4antirez.com 2nd of 3 in this piece