qertoip 💾

We live in the most exciting times ever!

[qertoip 💾 @qertoip]

@willmcgugan You will like https://t.co/igNIdqcV2T

[qertoip 💾 @qertoip]

PT and ES are slow to fix energy outage because they have no power

[qertoip 💾 @qertoip]

Next: please add tree-sitter based code excerpt. Modules, classes, functions signatures, etc.

[qertoip 💾 @qertoip]

Yes @cursor_ai thank you and long overdue BTW!

[qertoip 💾 @qertoip]

@burkov Plus, there is indeterminism on top of that, where the very same model scored a task perfectly yesterday and fails today, given exact same prompt (no memory).

[qertoip 💾 @qertoip]

@burkov Dark mode book.

[qertoip 💾 @qertoip]

#VibeCoding will age badly (but never bet against AI progress).

[qertoip 💾 @qertoip]

We need Extreme Programming 2.0 and Agile Manifesto 2.0 to guide AI-native reality of software engineering.

#ExtremeProgramming #Agile

[qertoip 💾 @qertoip]

GPT-4.1 GPT-4.1-mini GPT-4.1-nano - 3 new models released by @OpenAI - API only for now.

Seems to sit in between Gemini 2.5 Pro and GPT 4o as per early benchmarks.

https://t.co/tk9tM41E6F

[qertoip 💾 @qertoip]

Hey @realDonaldTrump, ask North Korea to confirm if tariffs exploded the local economy. They've been running this very-limited-import experiment for many years now - they should have reliable data and demonstrable results.

[qertoip 💾 @qertoip]

I started a new X channel "AI Pure Signal" https://t.co/k0EG6UQw47 - zero noise or distractions - just consequential AI developments as I see them unfolding.

Like & subscribe as they say ;-)

[qertoip 💾 @qertoip]

Excel, no code, and vibe coding belong to one family.

They allow people to duct-tape business processes and satisfy local goals without dealing with IT.

[qertoip 💾 @qertoip]

April 2025 Technology Radar published:
https://t.co/ISuP3eLJvr

[qertoip 💾 @qertoip]

Vibe coding AI will pick the most popular technologies like JavaScript, React, MongoDB and their ~2-4 years old versions.

You can prompt your way out of this but it will be increasingly uphill battle the further away you go.

This is a very unfortunate dynamic.

[qertoip 💾 @qertoip]

#LLM-s doing so great in coding is in part indication we were not as inventive as we believed we were as programmers. We might have been glorified hammer men to an extent ;-)

[qertoip 💾 @qertoip]

@simonw I do not. I completely forgot it's "year of the agents".

[qertoip 💾 @qertoip]

Can state-of-the-art #LLM-s read a clock?

https://t.co/yp2rRspdxz

[qertoip 💾 @qertoip]

@fchollet It's true but there is a twist: if model invents a cure for cancer no human could invent we don't really care it burned $1M in compute.

[qertoip 💾 @qertoip]

@arcprize (for real though - the challenge can only discard but not confirm any model to be an AGI)

[qertoip 💾 @qertoip]

The new ARC-AGI-2 has the same form factor of 2D color puzzles but goes beyond visuals to higher-level semantics and physics. I confirmed myself to be AGI ;-) by solving a few puzzles: https://t.co/GlITMx7Tcg cc: @arcprize