The second dimension to explore vs exploit is the faithful vs the corrupt. A faithful explorer believes in the unknown; that the universe is unfolding towards something better. The corrupt exploiter has no faith in the unknown. They'll sacrifice the young for the old.

"Wow this is fun. Wait am I building a query optimizer?"

Has anyone tried aligning dating apps incentives with an escrow contract where all parties only get deposits back or paid for the service when desired outcomes are met (dates, relationships, etc)?

How it feels telling an agent to implement the plan:

Agents are not good at finding concurrency bugs. To be fair humans aren't either.

This coding agent keeps hanging and giving up implementing a causally aware lineage classification and hybrid rehydration workflow checkpointing strategy. I paid you $3.35 and I expect results! The audacity.

I don't want an AI Agent in my browser because its only a matter of time before it notices how many tabs I have.

Everyday my life gets closer and closer to this:

Elixir folks dunking on the Rivet guy but at the same time I don't see us shipping a high level durable actor model product at a similar abstraction when we have all the means to do so. We could be giving the BEAM's runtime capability to all languages if we want. Rivet is neat.

Hi, friends! In the past 6 months Dashbit helped fund a part-time contributor for Numerical Elixir through Erlang Ecosystem Foundation. We're nearing Nx 1.0 and are looking to extend the funding! If you or your company use Nx in production, please consider contributing! every.org/erlang-ecosyst…

Alpha release of github.com/zblanco/runic, a tool I've had cooking for way too long. Will be writing / posting more about it in coming weeks.

Me to Codex: "make simple additive changes as described in this plan to prove the approach works narrowly so we don't over-complicate anything" Codex: "TIME TO OVERCOMPLICATE"

GPT-5.4 puts background colors and borders on *everything*.

The reason everyone is building the same prompt interface into every product is because its the simplest UX that works until we have a brain computer interface and even then its the same command interpretation UX just skipping natural language.

If you're afraid of subsidized tokens going away you just have to use fomo to gaslight as many as possible into subscribing so your usage is subsidized. Basically gym membership economics.

The bottleneck to utilizing human labor potential (coordination) is in effectively modeling merit / competency of an individual at scale. This is largely a high dimensional data problem and something AI can solve but only if identity is solved first.

What are the collective consciousness requirements to safely proliferate a given technological capability? If a capability is going to proliferate as a given and individuals or groups exist that would likely use it catastrophically what is the best intervention?

Last few years I've been feeling more and more bearish on "x for Elixir" projects and more bullish on "give Elixir superpowers to everyone" projects.