Augustinas Malinauskas

PhD Theoretical Physics Oxford

[Augustinas Malinauskas @amgauge]

https://t.co/W3pAKeKrE7

[Augustinas Malinauskas @amgauge]

@Keir_Starmer @grok is the decision based on rigorous evidence with causation determined beyond the reasonable doubt or poor data and poor interpretation of the data?

[Augustinas Malinauskas @amgauge]

@iamtrask You need to prove performance with coding agents not just deep research tasks

[Augustinas Malinauskas @amgauge]

@FGuzmanAI @karpathy Very cool! How long did it took to do it and how much it cost?

[Augustinas Malinauskas @amgauge]

@pushmatrix @mldotink can do it

[Augustinas Malinauskas @amgauge]

@ship_temp_md In my experiments database didn't work using Sites yet, but I imagine they will extend the functionality. In this case communication with DB would happen on server side.

[Augustinas Malinauskas @amgauge]

https://t.co/W3pAKeKrE7

[Augustinas Malinauskas @amgauge]

https://t.co/jQnxOUznmt

[Augustinas Malinauskas @amgauge]

The next step function happens by turning hardware into software

[Augustinas Malinauskas @amgauge]

@hthieblot Post-vibecoding

[Augustinas Malinauskas @amgauge]

@Lindyydrope Good point, @mldotink is best suited to handle this. If you have your agent in github or dockerfile Ink can deploy as many as you want

[Augustinas Malinauskas @amgauge]

@Patriot5715 @PhysInHistory @grok objectively evaluate this post. Be scientific and factual. Give receips

[Augustinas Malinauskas @amgauge]

I was thinking about this quote today too. You can delegate execution to AI or you can delegate both execution and understanding. For small ephemeral projects it’s fine to delegate both and never not worry about it, but when building complex projects you can’t delegate understanding yet, it will end up being a mess

[Augustinas Malinauskas @amgauge]

@rezoundous 6. Your mom
7. Grok

[Augustinas Malinauskas @amgauge]

When you look at a chair, what you really see is light that bounced off it and landed in your eye. When you listen to a song, what you really hear is air trembling against your eardrum. When you touch someone’s hand, what you really feel is your own nerves firing, never the hand itself.

The universe itself is dark and silent. Color, music, the warmth of someone’s skin — none of it exists out there. Your brain invents all of it, in the dark, just to make the data worth feeling.

[Augustinas Malinauskas @amgauge]

What is real or not is philosophical question. In standard model elementary particle is defined as irreducible representation of the Poincare group. It’s quite mathematical because connecting it to everyday concepts (like a little ball) leads to flawed intuition about what’s actually happening. Electron is elementary in the framework of the standard model. It’s neither particle nor a wave, it acts in its own unique way. We interact with the world through sensors, we don’t know that the universe really is, we only get to experience post proccessed signals that come to out brain. Particle or a wave are constructs that we invented, but we don’t know what nature truly is.

[Augustinas Malinauskas @amgauge]

Does @opencode has a benchmark how it performs vs codex (both using gpt-5.5)? @thdxr

[Augustinas Malinauskas @amgauge]

@johnny_makes can you share more about token consumption for each architecture of the eval

[Augustinas Malinauskas @amgauge]

To minimize human attention spend orchestration has to be delegated to a system that can capture your taste. OpenClaw+memory is good direction, but the implementation it not good enough yet. I think really good memory capturing taste, preferences and facts, agents reviewing each other and right UI could be big unblock, but I haven't seen such product yet

[Augustinas Malinauskas @amgauge]

@axtalks 107