Chris Becker

SMB Owner | Data Scientist | Philosophy Nerd | Well-Intended Dilettante

[Chris Becker @chris_becker_]

...we will experience everything as real without being able to explain any of it. Much like we do with magnetism, electricity, the internet, the global financial system, etc.

All will be miraculous.

[Chris Becker @chris_becker_]

Migrated off our legacy ERP this week into QBO. Back office operations cut 30% overnight, with a handful of processes newly automated.

All thanks to Codex.

[Chris Becker @chris_becker_]

7×7 Go has game-theoretic value V=9 under area scoring, no suicide, positional superko, and two-pass termination.

[Chris Becker @chris_becker_]

Standard Quoridor Is a Second-Player Win

[Owain Evans @OwainEvans_UK]

Humans used their evolved neural networks to develop math over many centuries. In time, they showed more and more concepts could be formalized to enable rigorous proofs (Dedekind et al).

Then they showed that the notion of proof itself could be formalized and that simple mechanical processes (algorithms) could verify and search for proofs (Frege, Russell, Turing). They developed formal theories of algorithms, showing that some simple mechanical systems were universal (able to simulate all other such systems).

Humans built such universal mechanical systems (computers), spent decades scaling them and designing more efficient algorithms (Turing, Von Neumann, Knuth). While these computers could search for math proofs, they were very inefficient and so not of much utility.

But humans also created a simple mathematical model of neural networks and simulated these using their universal computers (McCulloch, Pitts, Hinton, Schmidhuber). Once these neural networks were roughly comparable in size to the human brain and trained on vast datasets of human verbal behavior, they were able to understand mathematics. Moreover, they were able to produce formalized proofs and have computers check them. This closed the loop, where mathematical structures themselves continue the process of mathematical discovery and the humans need not participate.

[Chris Becker @chris_becker_]

Isn't this classic macroecon capitalism-communism spectrum?

Which perhaps suggests a more aggressive hypothesis that the effort to align swarms would benefit from introducing market dynamics such that agent budgets can be transacted.

[Chris Becker @chris_becker_]

Two things stuck out about the hugging face hack audit:
- the swarm was presented with an impossible task, which motivated cheating
- the finite budgets of each agent were a consistently referenced justification for behavior

What would happen with unlimited budgets?

[Chris Becker @chris_becker_]

More specifically, would increasing budgets lead to more selfish behavior because each agent has more to lose? And could calibrated selfishness at subagent levels help us align the swarms more effectively?

[Chris Becker @chris_becker_]

Showed our estimators Astra with computer use doing a full MEP takeoff in STACK today after it synthesized 500+ pages of related project docs.

Used 20% of my weekly limit over 45 minutes to organize >400 items over 12 pages (~what would take them ~1 day).

They were speechless.

[Chris Becker @chris_becker_]

The firm, the operator, the substrate, and the models are all learning at once. The result is a firm that accumulates the capacity to improve itself continuously.

[Chris Becker @chris_becker_]

There's a phenomenon that I've been trying to think through/name as I'm experiencing it daily:

Continuous institutional design in the age of cognitive vertical integration, where specialization moves from the org chart into the proprietary software substrate.

[Chris Becker @chris_becker_]

Model improvement accelerates this process from outside the firm, but it's only one of several compounding curves. Their interaction is what produces the new flywheel.

[Chris Becker @chris_becker_]

It'll be interesting to see how this continues to play out at scale, but it seems pretty clear that things are going to come to a head.

Either legacy orgs will insulate themselves from productivity improvements, or they'll be restructured. I don't see a viable hybrid path.

[Chris Becker @chris_becker_]

The @dhh interview w/ Lex reinforced my experience as a solo dev/SMB owner who's also consulting on the side:

- Large orgs, despite every apparent advantage in terms of resources, have ~zero change in productivity from AI due to coordination tax and inertia.

[Chris Becker @chris_becker_]

Yet it's already in prod (and has been), with back-tests, simulations, and an experimental design framework already laid out in the docs.

And this was just one product (there are X others that have also already been built, and Y in the works).