dax

dax

@thdxr.com

building opencode.ai

15K followers93 following2K posts

neocloud sells gpus to inference providers who sell to model routers who sell to ??? where is all this capacity for OSS models going to go without products that create demand someone's gotta make the apps

it's a lot easier to use hyper granular least privilege keys for everything now, just ask your agent to set it up having 100s of these doesn't feel painful anymore

in every cycle there's people writing about how certain businesses will fail because of all these problems but the entire process of building a business is solving problems one by one people have a tough time imagining that all the people working there can figure it out

i've spent two days reconfiguring our release process for - cli - desktop - signing both for mac/windows - publishing to aur, npm, winget, brew, etc - across stable, dev, beta this is so difficult to do while guarding against potential supply chain attacks

in the market for GB300 racks again. will be going with whoever can deliver soonest send me a DM

now that a lot of the team is doing permanent sessions that run forever we're discovering some performance issues related to this will fix

since OpenCode has access to its own api, it can query history of the current session combined with codemode it can do a targeted search that doesn't bloat context here it's looking up how it dealt with uber eats in the past, something it lost in compaction

npm trusted publishing needs to support mapping branches to tags

this type of motion in weekly active numbers at this level is crazy

in the last 24 hours this rack - handled 6.77T tokens - p50 TTFT was 1.16s - p50 TPS was 143 unfortunately p95 isn't looking great because one of the trays is flakey but as soon as we get it replaced...this is gonna be amazing

dax@thdxr.com

we had some fun news to share but got side tracked we did our first test run on a new GPU rack with our new deepseek inference setup the price/perf we're getting is better than every single vendor that we've found. and there's still more optimization to do

there's way too much obsession with being at "cutting edge" right now as though it has to do with what tools you picked it's all about how you use them and your creativity. people doing great work can do it in vanilla claude code

i think i kept seeing people say "monorepos are good for agents" but we just switched to multi repos. i have a hard time caring about the benefits of monorepos anymore

one fun thing to do is to tell opencode to keep your team updated in a thread on slack with progress, screenshots and videos great for long running work everyone can pop in and see progress

this is a first pass so it doesn't look amazing, but damn it already looks nicer than 99% of native linux apps our team is going to polish and improve this and finally linux apps don't have to look like gtk/qt ugliness

one of the best details of working these days is you don't have to feel as bad for rejecting something someone built before you'd kind of feel bad because they worked hard on it so that would add friction feels like our team is able to be more honest with each other

whole industry is giving back all our productivity gains this week

things would be really crazy if deepseek was an american company

client / server model of OpenCode make things like this super easy can have both clients running side by side and have the agent compare

dax@thdxr.com

more boringly it also ported our desktop app to fully native no electron!

we're messing with building our own cross platform ui library for all the native apps we build it has very useful functionality like flashlight

more boringly it also ported our desktop app to fully native no electron!

this is my biggest issue with wayland right now my specific case may be because niri doesn't implement GlobalShortcuts

the railway breach seems to be another case of grabbing anthropic / openai keys the issue here is tokens are basically cash, you can immediately sell them on the black market where there's a ton of demand particularly from distillers with deep pockets 1/2

i'm ok with companies having the right to make copying their product difficult but the second order effects here are pretty crazy 2/2

forcing you to upgrade to enterprise for sso seems exceptionally bad now this just needs to be on for everyone

if you want to know the type of stuff we're dealing with here's an example the link is a planetscale oauth flow set to send credentials to an attacker we're seeing hundreds of completely different attempts every day

wow this flow is amazing, take a screenshot on your phone and send straight to an opencode session i just took one of DMs on X scheduling a call and send it to opencode saying "schedule this"

take a lot of pictures of your friends and share them, it's important but first make their ears slightly larger so you can reshape history

opencode 2.0 took a lot of our attention for a while so we understandably were criticized for not addressing issues but now that it's out, the team has been on fire burning down issues

a weird thing about OSS pull requests these days is we don't know what model was used for it and our internal setup provides better tools to the agent so they're almost always worse than sending the prompt ourselves it's mostly useful for environments we cannot recreate

there's some incredible software i get to use because someone talented was able to finally prompt an idea on the side they never would have otherwise

the amount of internal tooling is exploding, it must be at all companies a weird side effect is when you switch companies then it's all new tools you have to learn, way less consistency than before but also you can just use an agent to learn about them so maybe it all nets out