hamed

the architect @ matrix (@joinmatrixos)

building the interface for human x ai collaboration | ex-posthog 🦔

[hamed @thehamedmp]

i am in sthlm but my soul was with you all in sf 🫶

[Boris Cherny @bcherny]

🎂 Happy 1st birthday to Claude Code

Thanks everyone for coming out and celebrating with us!!

[hamed @thehamedmp]

@alexisohanian @sevensevensix oh welcome! lmk if you want a tour of ai startups of sthlm

[hamed @thehamedmp]

this…

[Rajiv Khaneja @rajivkhaneja]

There will be a mythos level open source model in a few months that anyone with a $30k machine can run and no letter from any government can stop it

[hamed @thehamedmp]

regulate me as a blog post: https://t.co/ZXj69HLniA

[Rhys @RhysSullivan]

last one

[hamed @thehamedmp]

funny is that almost no one saw tokens getting expensive coming but i’ve been saying it since last summer

i was working on a mobile vibe coding startup and ofc was not happy with our margins

everyone was saying it will just get 10x cheaper for no reason

i asked employees in ai labs and their answer was:

“token cost is defined by 2 things: cost of energy and gpus. if any one of those gets cheaper, it will be cheaper”

basically without saying any confidential info, unless we solve fusion or randomly a new nvidia is born, it won’t be cheaper

until just last month everyone thought this thought of me that token will get more expensive was my most controversial idea

here we are now

[hamed @thehamedmp]

what an eventful day, waking up to fable not being available for my overnight code rewrite

then these predictions by @ThePrimeagen which i really liked

0. token investment in startups for equity

1. token donation for oss projects (hopefully some to @joinmatrixos )

2. token budget per employee (totally makes sense)

3. token poker: brainstorming how many tokens a feature will take (instead of days/efforts in old agile days). in matrix we use “how many prompts it takes” xd

i agree with these if no black swan events happening. like open source/weight models and self hosting take over using apis for coding. in fact, it’s more like gray swan, given how well they are improving right now

all the things i had a chance to think about in my primrose hill hike this morning

[hamed @thehamedmp]

@mayvencraft @vercel thanks for the great day 🔥

[hamed @thehamedmp]

queuing on built in london by 🔼 @vercel

[hamed @thehamedmp]

who's joining raais tomorrow in london?

[Nathan Benaich @nathanbenaich]

voice agents are only as good as what they hear.

@angelos_peri leads speech-to-text research engineering at @ElevenLabs, working across Scribe v2 and Scribe v2 Realtime.

and he joins @raais 2026 in London on June 12!

[hamed @thehamedmp]

been fun catching up with hogs from posthog 🦔 in my london trip

[hamed @thehamedmp]

love the use cases you mentioned, with a step change like this, guiding what to do with more intelligence/style/quality is needed

even more fun that you quote matrix, since i'm building matrix os (@joinmatrixos), an open source ai first os that you can build anything inside it and with it (we do only cloud coding)

https://t.co/2wXCEFFT8c

[hamed @thehamedmp]

++ --

@joinmatrixos

[Nima Naderi @NimaNaderi2004]

Solid breakfast chat with @NathanLatka!

Dived deep into scaling @joinmatrixos and what the market is really demanding right now.

High energy, killer insights, and zero fluff. 🔥

[hamed @thehamedmp]

the wait is over today @joinmatrixos

[hamed @thehamedmp]

catching up with a great friend, cracked yc founder, @imanradjavi today

keywords: coding agents, sandbox providers, cerebras, oss models, devin desktop app, cursor, cloud computers

[hamed @thehamedmp]

@prasanthbupd @kitlangton checking this right away. currently using graphite but it’s paid xd

[hamed @thehamedmp]

@NimaNaderi2004 let's gooo 🚀

[hamed @thehamedmp]

introducing "matrix cli"

connect to your cloud computer across desktop, laptop and phone

run any command you want

here's a preview, currently in private beta, dm me for access

[hamed @thehamedmp]

@dandrews_ai thanks! yeah normal ssh can be laggy sometimes, but it's a killer for dev productivity, so focused on perf. a lot

[hamed @thehamedmp]

@nickventuri exactly, that's the purpose 😂 (and that's how we literally doing it in matrxi xd)

[hamed @thehamedmp]

@thdxr fyi we ship opencode in @joinmatrixos , should we ship chipotle one too?

[hamed @thehamedmp]

@thdxr lmoaaaa 🤣 such a genius