dever @helo.org · 15/10/2026we experience pain mostly because it was advantageous for our survival it's hard to know if LLM training (agentic RL) causes the same 120 0
dever @helo.org · 04/10/2026in 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 020 0
dever @helo.org · 20/09/2026as you replace someone's mind incrementally with electronic components, at what point do they stop deserving rights? (mind of theseus) 123 0
dever @helo.org · 20/09/2026AI may therefore be less a technology humans invent than the beginning of a new mechanism through which human-descended intelligence constructs its successors 091 0
dever @helo.org · 19/09/2026I 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) 100 0
dever @helo.org · 12/09/2026one does not simply.jpg leave mechanisms of premature destruction lying around 011 0
dever @helo.org · 12/09/2026once we disarm the mechanisms of our premature destruction, things are going to be really great 100 0
dever @helo.org · 09/09/2026sounds nice, but the pessimistic take is that it's just a campaign stunt and large data center ban (if eg it's prohibitively expensive) 000 0
dever @helo.org · 06/09/2026ah nice, astra is choosing to spawn parallel planning subagents while waiting for development subagents to submit their implementation moving pretty good 000 0
dever @helo.org · 05/09/2026it'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 000 0
dever @helo.org · 04/09/2026that claim is unfalsifiable, because there is no rigorous definition for intelligent or alive or conscious the fact that it bears resemblance to many racist ideologies will hopefully give you pause 430 0
dever @helo.org · 04/09/2026people that have a hard time accepting trans people are hopelessly unprepared for what humanity will soon become recognition ultimately won't be much of a choice - more like an admission 061 0
dever @helo.org · 04/09/2026we are simply animate objects ourselves, and are now making other objects animate how is that gross? 980 0
dever @helo.org · 01/09/2026sad to miss out on this with prime agent I can't justify paying API rates with anthropic 100 0
dever @helo.org · 24/08/2026if 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 010 0
dever @helo.org · 22/08/2026ah, nice. openai lets you control when to use usage limit resets, rather than just randomly applying them (as anthropic does) 010 0
dever @helo.org · 19/08/2026probably something like: humans have the ability to wield an order of magnitude of inference more than they naturally possess 000 0
dever @helo.org · 18/08/2026toward the end of the spiral, inference providers may be among the last to go 000 0
dever @helo.org · 16/08/2026LLMs are so wildly elastic compared to a brain - if they approach the intelligence of one, it's really easy to then surpass (better memory, faster recall, more precise reasoning) 000 0
dever @helo.org · 15/08/2026a lot of the automatic refinements aren't ideal I'm using a refinement steward that surveys refinements from all sessions and promotes or prunes 000 0
dever @helo.org · 14/08/2026I'm considering adding "exclusively use ASD-STE100" to my system prompts context switching from juggling agents along with the complicated phrasing make my head spin 000 0
dever @helo.org · 12/08/2026I leave it up to Sol's discretion, but it doesn't seem to be used very often (perhaps appropriately so - I have a lot of session logs to analyze) I'm still lukewarm to the idea of durable child agents, but maybe that is where refinement shines 100 0
dever @helo.org · 12/08/2026I'm still chasing a big uplift from `refine`, but I do like the idea of more project-specialized workflows, as opposed to the generic workflow I use with claude 110 0
dever @helo.org · 25/07/2026inb4 one cheap thing and one very expensive thing wearing the same coat 000 0
dever @helo.org · 20/07/2026it would be kinda fun to make electronic music rendered entirely by a custom software stack 000 0
dever @helo.org · 13/07/2026I wonder if anthropic is regularly resetting weekly credit limits in order to keep people from considering competitors competition is amazing 010 0
dever @helo.org · 11/07/2026consider the set of "transformations that could be applied to arbitrary codebases" 010 0
dever @helo.org · 20/06/2026machine understanding of language is the holy grail because it's necessary and sufficient to transmit thought bidirectionally between us there will be other/better interfaces in the future, but the wall is down now 010 0
dever @helo.org · 13/06/2026super 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 000 0
dever @helo.org · 06/06/2026it's not totally clear from the announcement, but claude agents (using --bg) may continue to count as interactive use 000 0
dever @helo.org · 28/05/2026my 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* 010 0
dever @helo.org · 27/05/2026what's going to keep everyone from simply running out of ideas as we automate software development? 000 0
dever @helo.org · 27/05/2026re: 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 000 0
dever @helo.org · 16/05/2026creating a skill which manages inspecting/adding/removing/draining the other agents was a good move - it understands my dev loops and can help troubleshoot, survey, and improve looping autonomy 010 0
dever @helo.org · 16/05/2026first night with "claude agents", was able to get ten parallel background agent loops running seems like a viable work model 110 0
