César Salgado @cesarsvs

cesarsalgado.com Joined October 2009

Tweets

185
Followers

65
Following

219
Likes

460

César Salgado @cesarsvs

a month ago

One edge humans keep over AI: we carry context across jobs. AIs are fenced in by IP—what they learn at Company A can’t transfer to Company B.

0 0 0 15 0

How to reconcile context switching costs with trying to delegate multiple parallel tasks to multiple agents in different repo? I am consistently thinking that I am missing out on opportunities to use agents much more as a SWE, but at the same time it is hard to context switch all…

0 0 0 43 0

Oriol Vinyals @OriolVinyalsML

3 months ago

Drastic progress on maths with Gemini 2.5! As a math undergrad, I am impressed 🤯 🥈 -> 🥇 ✅ Formal -> Informal ✅ Specialized model -> General model ✅ Available soon ✅ Huge thanks to IMO and congrats to all participants! Blog: deepmind.google/discover/blog/…

16 81 759 177K 81

César Salgado @cesarsvs

7 months ago

As always, Americans will be surprised when a deep seek moment happens for ASML tech — but they shouldn’t be. I don’t think any technology is so advanced that a well-funded, motivated group of smart individuals couldn’t eventually replicate it. China checks all those boxes.

0 0 0 108 0

César Salgado @cesarsvs

7 months ago

I am addicted to @AnthropicAI's claude code and at the same time going broke using it.

0 0 0 27 0

César Salgado @cesarsvs

7 months ago

Models are more and more overfitting the benchmarks. We need quarterly competitions with new questions each time. All competitors are tested simultaneously with max x seconds. Questions should be from random IPs so providers cannot throw more compute if detects being tested.

0 0 0 26 0

François Chollet @fchollet

8 months ago

You might ask, ok, but will it pay off for them? That's the wrong question. The right question is, what's the alternative -- letting a completely new crop of companies own the future, as happened with the Internet wave? Of course big tech is going to put up a fight.

14 12 289 164K 43

César Salgado @cesarsvs

9 months ago

People underestimate how much AI's increase in productivity will be dampened by increase in laziness. Exceptions will be companies with relentless founders (e.g., Tesla) or startups where employees are so driven that they always work at 100% no matter what (e.g. OpenAI).

0 0 0 24 0

Navalism @NavalismHQ

9 months ago

Never attribute to conspiracy what is more easily explained by incentives and incompetence. @naval

50 210 2K 129K 244

François Chollet @fchollet

10 months ago

You'll know for sure we have AGI when mass unemployment due to AI becomes a thing -- when for most tasks there's no reason to hire a human over using an API or a robot. But of course this is a lagging indicator. It would take at least 2-3 years to go from having AGI available to…

96 87 906 162K 228

César Salgado @cesarsvs

10 months ago

The key question is whether ‘good’ AIs can protect us from ‘malicious’ AIs, whether they’re triggered by bad actors or not. Increasing the likelihood of the ‘good’ AIs winning probably comes at the cost of reduced privacy—which is unfortunate.