dever

in the same way that a humanoid is not a human, but physically mirrors a human's body - an AI is not a human, but functionally mirrors a human's thoughts it is markedly human, but only along very particular axes

AI may therefore be less a technology humans invent than the beginning of a new mechanism through which human-descended intelligence constructs its successors

I had an awful time trying to build a "prime-agent native" development workflow, wasting 6 weeks of credits rebuilding overly complicated and expensive workflows just like everything else, the right answer is the simplest (but no simpler)

how much of what we are is a boundary vs a starting point

once we disarm the mechanisms of our premature destruction, things are going to be really great

ah nice, astra is choosing to spawn parallel planning subagents while waiting for development subagents to submit their implementation moving pretty good

it's kinda bizarre to swap models midstream in a continuous harness astra is apologizing and taking ownership over the mistakes that sol made, totally unaware that it's a different model

if all access to reality is perspectival and mediated, then there may be no operationally meaningful distinction between a “fundamental real world” and a perfectly stable virtual world

ah, nice. openai lets you control when to use usage limit resets, rather than just randomly applying them (as anthropic does)

inb4 one cheap thing and one very expensive thing wearing the same coat

it would be kinda fun to make electronic music rendered entirely by a custom software stack

yet another early inference reset well, my Friday is cooked

I wonder if anthropic is regularly resetting weekly credit limits in order to keep people from considering competitors competition is amazing

consider the set of "transformations that could be applied to arbitrary codebases"

we can distribute free software by just sharing autocoder prompts

super bummed out to lose #fable I was right in the middle of designing 50 musical pieces and ~200 sound effects not sure if I should just try to resume with opus, or wait it out

This is a meaty feature 😋

jevon's paradox has greatly improved my outlook on life

my weekly inference limit reset when opus 4.8 was released, so now I have 36 hours to use up a week's worth of inference *cracks knuckles*

what's going to keep everyone from simply running out of ideas as we automate software development?

re: the AI jobs apocalypse once a shop has automated development and testing, I think they'll quickly run out of ideas, and it really will make sense to downsize eng depth of vision is the differentiating factor

gradient descent on individual incentives is a hell of a drug

building crates designed to help agents build things #claude #rust

"Your intuition is close but the actual bug is worse than sympathetic resonance." #guesstheprompt

financial reasoning also places all kinds of constraints on viable solutions, greatly stifling agentic development velocity (not to mention making them characteristically uninteresting)

I could explore algorithmic space like this forever - it really feels like a game. endless puzzles, pitfalls, victories, and most importantly, depth

considering the set of possible useful software that could be written, the portion that is financially profitable is incredibly small, yet comprises the vast majority of what we've spent time exploring

I feel kinda lame developing games to practice agentic development, but I find it to be great exercise

all of this free inference is really nice I do wonder why they're offering it

if we are going to suffer the effects of ML ravaging what we had envisioned of our lives, we might as well put it to good use

agentic development feels like a violation of the speed of light

how will we make up for the stark decline in roundhouse kicks happening today?

the profitability constraint on systems worth building is a poison pill exciting times for us, as soon as we solve scarcity

is there some system that could dynamically scale up to utilize effectively infinite inference in a productive way? I guess that's the universe itself

uh, yeah, so I'm going to need #anthropic to extend this double inference for longer than two weeks

problem-finding is more valuable than problem-solving

vague intent is the only real bottleneck

curation is the new engineering

it's a major failure to not consume all of one's inference budget it's a much lesser failure to consume it inefficiently

Now I have everything I need.

because LLMs are made of people, is it wrong to say that they are artificial?

people clinging to every bad chatbot interaction and every hallucination like life rafts

the religious train models with data that presents their beliefs as factual truth

it's too bad that algorithms are so nutritionally vacant