dax
thdxr.com
dax
@thdxr.com
building opencode.ai
every shortage ever has been met with a glut soon after
this better happen
there just aren't enough GPUs it's crazy that even a company our size is being bottlenecked by this
February 13, 2026 at 9:21 PM
there just aren't enough GPUs it's crazy that even a company our size is being bottlenecked by this
February 13, 2026 at 9:00 PM
someone give me a waymo invite code
February 13, 2026 at 5:28 PM
i think LLM bill per employee is now higher than health insurance

maybe i'll make people pick one or the other
February 13, 2026 at 5:13 PM
if you manage a team how would you feel in a scenario where the weakest member is spending the most on LLMs

like they're spending $2k a month or something
February 13, 2026 at 5:12 PM
it's crazy how good our brains are at noticing patterns

we all instantly pick up on "that's not x — it's y" as llm speak without ever consciously registering this pattern or being taught it
February 13, 2026 at 2:16 PM
i've always been pro making money, being money motivated

and if i tried hard to do something great then money would be a side effect

it feels like silicon valley lately is infected by a different version of this where the money is the entire point

very uninspiring
February 13, 2026 at 1:56 PM
a lot of companies know they should be doing something in ai but haven't found clarity yet

which is why every new concept that comes out gets over developed

these companies jump in and build something to try and grab some territory but usually it doesn't make sense
February 13, 2026 at 1:16 PM
a lot of our recent hires care deeply about code

how's it's written, how to make it elegant, how to make it easy to understand and a joy to maintain

it's a counter bet in a field where most people are saying we won't be looking at code for much longer
February 13, 2026 at 1:20 AM
everything takes forever
February 12, 2026 at 11:19 PM
so far we've tried to upgrade opencode to bun
1.3.6
1.3.7
1.3.8
1.3.9

even with a beta period each time it's caused a huge number of crashes for users so we keep rolling back to 1.3.5

sorry for the churn we're probably going to give up on this for a while
February 12, 2026 at 4:45 PM
interesting thing about minimax 2.5 is it's a smaller model

considering it's very usable it's a great candidate for home labs

also would love to see inference providers try and max out its tokens/s can probably do something crazy
February 12, 2026 at 4:30 PM
all the oss models are trying to be an opus killer

not so much a codex killer

this makes sense because opus is more addicting to use so that's what's worth competing against

but wonder if eventually someone takes a swing at a codex like model
February 12, 2026 at 4:16 PM
minimax 2.5 is now generally available and free for 7 days in opencode

i'm going to try and switch to it as my default so i can get a sense of how it works

golden era for opensource models right now
February 12, 2026 at 4:15 PM
if ur team isn't gang prompting ur ngmi im talking 4 meat boys 1 clanker all in a thread hitting that bot with conflicting instructions until it spills its tokens all over the place
February 12, 2026 at 5:01 AM
someone give me a good argument as to why LLMs won't commodify

note this doesn't mean big labs don't continue to do well, just means they don't have monopoly power in it
February 12, 2026 at 2:43 AM
codex is by far a better coding model than opus - anyone who knows anything understands this

but the whole industry should reflect on why opus is the most popular

people assume whatever is the smartest will win but the old rules of product are still what determine everything
February 11, 2026 at 7:58 PM
so much of what's going on in ai from the labs being valued the most to browser automation to standards like MCP look so much like other things in history

and in those cases there was a dramatic flip

on one hand i am certain it'll happen again but it's so hard to imagine
February 11, 2026 at 7:56 PM
ai writes all our code for us*

*team chasing down the most fucked up bugs and perf issues cleaning up after my dumb ass not pictured
February 11, 2026 at 4:24 PM
we've taken the first steps in eliminating opencode's ripgrep dependency :)
February 11, 2026 at 3:22 AM
"x is too dangerous so i am the only one who can be trusted with it" is such a hilariously cliche trope it shows up constantly in history, books, movies

and it's the same outcome every time

and yet we're watching this happen in real time with AI
February 11, 2026 at 2:37 AM
raw LLM prices don't mean much but we can look at real world data to see what people actually spend

from last 7 days avg cost per session
February 11, 2026 at 12:55 AM
the whole TUI vs GUI debate no one makes the obvious point

why would you want to use something gooey

ew gross
February 10, 2026 at 8:14 PM
did everyone end up adopting .agents/skills convention i wasn't following
February 10, 2026 at 5:15 PM
wow we're getting a vps vs serverless moment in 2026 bringing back the classics
February 10, 2026 at 4:11 PM