CoFounder & CTO for https://xitter.cf/t.co/xJxyBek82O - deploying AI agents and SLMs in manufacturing Talk to me about NLP, ML, or squeezing the AI dry when doing hard work
New York
Joined November 2008
- Tweets1.4K
- Following697
- Followers286
- Likes861
what i love about rust is that it has first class support for functional programming. no type erasure.. yaay
verbose as it is, 300 lines of rust are packed with logic compared to 3K typescript or heaps of python
An entertaining reading that's insightful and quite revealing at the same time
Great essay from @clarabcollier: asteriskmag.com/issues/15/af….
(You should subscribe to @asteriskmgzn, in addition to @WorksInProgMag of course. They have a very good print edition.)
Looking forward to it
Excited to host this one with the @OpenAIDevs team
Will be there on both coats, come say hi
Victor Paraschiv retweeted
Grok 4.6 was the first model trained on internal model-development tasks.
My team built the training and evaluation stack that enabled Grok to learn from work that accelerates model development itself, including production inference and kernel optimization.
Grok 4.6 now leads our internal MTS Eval and InferenceEval. In one experiment, we gave an earlier checkpoint our production inference codebase in a fully automated environment. It explored 297 optimization ideas and shipped 3 changes on top of an already human-optimized stack, improving prefill throughput by 3.1% and decode throughput by 1.5%.
Our work is featured in the R&D Enablement section of the Grok 4.6 model card.
Anthropic and OpenAI must be doing the same thing to enhance model-harness affinity
Destroying books is equivalent to destroying art.
One must not destroy something it cannot create.
And they call themselves educated!? We're falling into dark ages, again?
I like that Cursor agent mode uses the selected model for top level work and it delegates to Compose 2.5 subagent work that do run the tool calls. Great value for money.
Axiom 6: Modernity dramatically reduced external constraints and created an unprecedented cognitive surplus of mature minds.
For most of history, disease, famine, war, and infant mortality dominated outcomes.
Today, billions reach cognitive maturity and live decades beyond it.
This changes the game. Thread 🧵
With external killers largely receded, the selection pressure has migrated.
Ideas, beliefs, and the quality of our thinking now have more room to compound, for better and for worse.
This is why critical thinking is no longer optional.
Full essay exploring all axioms and theorems → vicpara.substack.com/p/the-a…
What’s one realization or belief shift you’ve got now, in new “post-maturity” potential?
Axiom 1: Human cognitive hardware is largely unchanged across recorded history.
The Romans, ancient Egyptians, and our savannah ancestors ran on the same brain we have today, same capacity for reason, imagination, abstraction, and calculation.
The difference was never the hardware. Thread 🧵
In the age of cognitive surplus and technological leverage, this over-built brain is finally getting the chance to run at full capacity.
The question is: what software (beliefs) are we installing on it?
Full essay with all axioms & theorems here: vicpara.substack.com/p/the-a…
What’s one idea or belief you’ve changed in the last few years that noticeably improved your life?
Reply below 👇
Human history is not steady enlightenment.
It’s changing constraints on the human mind.
For the first time, survival and flourishing depend less on famine, plague or wars, and more on the quality of our ideas and thinking.
A new evolutionary regime is here. Thread 🧵
This is the core argument of my new essay: “Ideas Have Become Selection Pressures.”
Full axioms, theorems, and implications here: vicpara.substack.com/p/the-a…