korrents
Gergely Orosz

What Gergely Orosz thinks about LLMs

@gergely-orosz · 23 positions · 0 changes of mind

Writes The Pragmatic Engineer, the software-engineering newsletter, and wrote The Software Engineer's Guidebook. Formerly an engineering manager at Uber, and at Microsoft/Skype and Skyscanner before that.

Everything they publish, on ppll ↗

Gergely Orosz 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.

5 dated positions, 2026, in their own words. Our reading of what Gergely Orosz has said — not written or endorsed by them.

  1. Our readingUsing open models is the approach offering the biggest AI cost savings, followed by smart model routing.

    Their wordsTo answer the question posed in the header of this report, it’s apparent that using open models is indeed the approach offering the biggest savings, followed by smart model routing. Spending controls and context optimization also bear down on costs, but they don’t come close to the first two techniques in results.

    ↗The Pulse: tech companies move to open AI modelsblog.pragmaticengineer.com

  2. 1 day earlier

  3. Our readingThe largest tech companies are not using fewer tokens; they are switching to cheaper, still capable open models instead of SOTA models.

    Their wordsthe largest tech companies (at the top 1%) aren't using fewer tokens, they're just being cost conscious and moving to cheaper, still very capable open models running vs SOTA model usage

    ↗@GergelyOrosz on Xx.com

  4. 1 day earlier

  5. Our readingExisting AI models were already more than good enough for writing code and do more than can be utilized from them.

    Their wordsthe existing ones were more than good enough, and I'm learning how much to trust those existing ones... and already they do more than I can utilize them (for writing code)

    ↗@GergelyOrosz on Xx.com

  6. 1 day earlier

  7. Our readingAn “AI trained on person X” cannot stand in for that person, because people change their minds in light of new information and the model does not.

    Their wordsThis is, by the way, also why the category of “an AI trained on person X” is absolutely BS. People change their minds in light of new information, the “AI trained on X’s writing/videos/tweets” does not

    ↗@GergelyOrosz on Xx.com

  8. 6 days earlier

  9. Our readingWriting is thinking, so an LLM inflating a one-sentence observation into three paragraphs destroys the thing that made it worth reading.

    Their wordsWriting is thinking, and we are interested in your own thoughts and observations, not a sharp, one-sentence observation turned into three paragraphs of blob text by an LLM.

    ↗How software engineering is changing: an essay challengeblog.pragmaticengineer.com

    writing