AI safety focuses on dramatic things like bioweapons and nukes but LLMs are breaking so many systems we rely on in more boring ways via spam, noise and misinformation wish this was more of a focus

a million posts on my feed telling me how to build agents zero people talking about building agents

yeah so it was actually like 2.5 months because i was doing that night shift thing with my newborn i guess there were some consequences

it's kinda crazy to think the world would be in a better place if a faceless corporation like google was leading ai

just got penetrated by mythos

since medicine is important if the FDA determines there's a shortage they loosen IP rules around it it allows other companies to fill gaps until the shortage is over (this is why there were all those GLP1 companies) 1/3

like i said yesterday this was already a thing since at least gpt 5.3 but now we have to suffer a cycle of confusing mystery and go through the whole "well it was bs last time but maybe this time its different" we're all just caught between these two companies

maybe gta6 is also too dangerous to release

there's an article floating around claiming OSS models could find the same vulns as mythos it's a bit confusing though - in this test they pointed it towards the problematic code and even very tiny models could find the problem 1/3

i'll post on here making fun of myself and people will get mad

no no i've definitely been doing a lot of work it's all just too dangerous to release

my retention so low call me boris churney

early on in tech cycles there appear to be a boom in a type of company so everyone builds things those types of companies will need but this is always temporary, ends up consolidating to a few players and there's no real market

the past decade has been all about figuring out how shitty software can be without drastically impacting revenue LLMs are mostly accelerating this search

when gpt 5.3 codex came out they had a new process because they were worried about cyber abuse we pass through user IDs to them so they can block bad actors easily - and we see a ton of blocks if only they thought to do a whole marketing video on it

guys i had a one of a kind idea for a business sandboxes for coding agents

opencode server now can run under nodejs as we're no longer using any bun specific apis we're working on a larger refactor now of its internals as we work towards a 2.0 will have nice updates to the plugin and sdk apis

there's gotta be one person at anthropic who uses opencode in secret

it seems we're in the phase where companies are polluting LLM training data with slop comparison articles against their competitors and just like with google if the LLM companies don't take it seriously it's going to make the product useless

people learn the following two terms and use them to sound smart 1. prompt caching 2. quantization breaking prompt cache is ok. you can't guess what will happen you have to look at data at scale from real users 1/2

i hope one day i get to realize all these unrealized gains

bunch of comments on HN saying i'm horrible/immature/petty to anthropic bunch of replies here telling me i'm "sucking their dick" mfs so amped up over a corporation that isn't gonna fuck them (at least not in a good way)

vite plus not clicking for me yet was hoping it would feel like bun for node can't fully articulate it but it doesn't feel quite right, def feels like something i have to learn

the world is so screwed

so my gen-z coworkers i noticed they say words wrong all the time or they'll mix up 2 similar sounding words is this a thing

engineers bias towards conservative approaches to business everyone loved when dario said he didn't want to overinvest because if he was wrong the whole business would die but you look at a lot successful companies and at some point they had to make the all or nothing bet

it is very hard to be both a platform and a product company it feels ok to compete with your users a little bit but you eventually run into problems you never could imagine ahead of time

how much can you jack up your salary if you quit cloudflare and join vercel and then quit vercel and join cloudflare in a loop

i've never ever seen a good external analysis of big startups so many writeups about openai or anthropic and all of them are basically entirely wrong happens with every generation of companies, why do people waste their time on producing this stuff

there are like 5 people in tech that understand marketing and of those 5 probably 0 have actually pulled it off, they just understand it correctly and are trying the right things so it's funny to see all the takes on TBQN acquisition, every single person is wrong

why has no one vibed up really good local otel tooling i want to be able to look at all these traces during local dev

some thoughts on figuring out how to keep producing good work

idk how people manage infrastructure anymore every service has their own bespoke cli / config file and they don't support terraform well anymore your system is never just one provider so do people just have a mess of these smashed together?

the amount of misinformation out there is crazy right now you can see this with the cc leak - people point their llm at it and it tells them some bullshit and they post about it with extreme confidence this is happening everywhere, impossible to keep up with and correct

there's always people arguing that most people can't notice quality to rationalize infinitely cutting corners they might not be able to articulate why but people can definitely feel when something is great

what if we gave you unlimited tokens for free and we also paid you

the hardest part about having a remote team is you can't hit your employees

1 month later 6,500,000 monthly actives

we will NOT be participating in the pagan holiday of april fools a day devoted to to ritualized deception and spiritual corrosion opens a pathway for the devil to do his work

you're going through the claude code source when the real action is going down at github/dmca

working on a new project and once again i do not understand how people use agents at the start of the project i didn't even think to, just naturally went around defining interfaces to think through what i'm doing agent comes later

claude code source is 512K lines opencode is 118K we're getting LOC mogged

do ai bots respond to posts with obscenities in it like if we end each post with fuck bitch asshole fuck will they still reply

people often ask if OpenCode Go will have a higher tier than $10/month we'd like to but let me explain what makes this complicated subscription plans need to offer more than spending the same amount on pay per token 1/4

we see a lot conspiracy theories claiming models on OpenCode Go ($10 plan) are served differently we're using the exact same providers you use when you go direct all providers are constantly tweaking things and sometimes there are bugs but it's pretty minor

if you're working on making something cool accessible to more people there will be a crowd of people who absolutely hate you

we all live in a big ecosystem that all feeds into each other - frontier labs push the bar on what's possible - opensource labs build off of that and make their own discoveries - inference providers make huge capex investments to serve these models 1/3

man you guys already got to all the good domains and npm packages and you don't do anything with em

am i the only one still on Xorg

right now everything in the world is telling you to go faster, ship more, add that feature, start another project so i'm actively working on feeling ok not doing any of that