V ✁ (@mislocating) header
V ✁ (@mislocating) profile picture

agentslop, startupslop

🩸[█░░░░░░░░░░░░░░░░░] 4% left
Joined March 2022
Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)Photo by V ✁ (@mislocating)
my bank's mobile app sends me crypto spam notifications every day in the same android notif category as legitimate transaction alerts but in one voice prompt now our cron equivalent automatically detects and clears them with a regex this is like the 10th general android ux improvement we've made this week with zero intervention from me needed
1
84
~1mo ago I was doing one agent thread per narrow task that gets archived automatically but Ive since pivoted to one long-term thread per project or subproject it's much better. now the project lead spawns sub agents for object level work and acts as a manager. then they're able to send and receive messages from the other lead threads to coordinate things across projects, eg for using and developing share infrastructure at this point even for the lead threads that report to and receive instructions from me directly, like 80% of their input messages are coming from other agents within and outside the project
42
nvm it was more like 1.5x, the skill cap is very high on this, I've realized efficient agent->human communication is what I'm researching and experimenting with. all of my agents are now exploring and exploiting and we take note of what I report as easily digestible
made a new UI and workflow for how agents communicate to me that reduces required time and attention by like 3x so that's effectively like 3x boost in work I can do, gg ez
38
my startup's bottleneck is how quickly I can breathe hot steam after a long voice prompt to heal my vocal folds and start the next one
53
made a new UI and workflow for how agents communicate to me that reduces required time and attention by like 3x so that's effectively like 3x boost in work I can do, gg ez
1
98
it's over I fell for the tailscale meme I'm voice prompting at the dentist now
1
1
108
some serious problems that were only fixed in the pre release yesterday so I'm not fully happy but yeah free security and remote access ux improvements should have trusted yacine when he said he wants to work there
31
must kind of suck to be theo and having half the population of earth be dependent on software that you also rely on for productivity it feels so good in my agent interface to see the slightest bug fix and just send off a voice prompt requesting a fix in 1 keybinding, having it auto commit and deploy and archive its own thread automatically afterwards it's never fully polished but also just grows very easily, like perpetually at 95% of what I want but with my desired scope increasing every day
51
sad to say this after doing basically everything possible through the terminal for many years... but after heavy workflow refactoring my computer is now basically an ipad, just listing agent threads in my t3-equivalent and bindings to start and stop voice prompts
1
1
108
for now many different projects require my visual judgment but imagine there was no monitor and no keyboard, you just worked through like a ring with tts and mic for voice prompting I guess you just stare at the wall or outside at the trees all day can't tell if it's dystopian or utopian
25
when codex sets wakeup alarms the self-prompts seem to spiral into a hypercompressed writing style? ofc the cot is like that but interesting that the philosophy applies I guess to any writing solely intended for itself
1
51
for longer like 10h+ tasks I've also seen it start doing this to general project documentation though
22
openrouter UX for the key security page is quite poor, almost impossible to horizontally scroll and see the multi-select ui but codex using cdp found it and did my config change easily, やっぱり another victory for DOM-based navigation. and it will be viable for yet a while since llm proficiency with the web stack is perpetuating its use
55
things have not been looking good for the portfolio since near left and stopped posting investment advice
1
54
andrej didn't mention this in his 2025 YC talk but lately I've been merging his software 1.0 and 3.0 in the form of fallback and automatic self-repair - you give input to program X - X tries processing it with its deterministic logic - if X detects a failure, it spawns an llm agent to examine the log (including the input) and path to X's src code - agent diagnoses the failure, implements a fix, while (!) preserving the running X process. such that the agent can then manually push the corrected post-fix output to the original user so from the pov of X's user (itself an llm agent usually), there was never any issue with X in the first place besides a longer running time. this is how I've set up all of my mcp servers to work
1
76
iterating on some reverse engineering with sol 6.1 for over a week and opus 5.5 more or less one-shot it to a better state than the final with sol lmao I need to improve my proxy benchmarks
1
1
84