Drew Houston

Founder/CEO of Dropbox · Board of directors at Meta

[Drew Houston @drewhouston]

@rileybrown Realistically you'd need two of these (384GB vram per node) -- quantized weights alone take up ~500GB VRAM: https://t.co/HztdJenSK7

Mac studios could work but will be really slow

[Drew Houston @drewhouston]

@mitchellh (2x blackwell tinyboxes can plug into 4 normal home/office 120V circuits, about the limit of "local"/non-datacenter)

[Drew Houston @drewhouston]

@mitchellh Looking like GLM 5.2 is truly Opus-tier -- to run it fast (>100 tok/sec) you'll need 8x RTX 6000 pros minimum ($125-150k), but achievable now

[Drew Houston @drewhouston]

@sriramk Congrats Sriram! Thank you for everything you did for 🇺🇸

[Drew Houston @drewhouston]

@spenserskates Thank you Spenser -- it's been awesome to see what you've built and I know you've been inspiring future members of the Battlecode mafia :)

[Drew Houston @drewhouston]

@sriramk Thank you Sriram! 🥂

[Drew Houston @drewhouston]

@schreier @Dropbox Thank you Bryan! Lucky to have had you on the journey and grateful for everything you did for me and Dropbox 🙏

[Drew Houston @drewhouston]

Today, we're promoting Ashraf Alkarmi to co-CEO of @Dropbox. Ashraf and I will jointly lead the company, and after a transition period, I'll move into the role of executive chairman and Ashraf will be sole CEO.

Ashraf has transformed our core business since joining — the business has gotten stronger every quarter under his leadership, and he's the leader I trust to run this company.

What’s next for me: my focus right now is making sure Dropbox is in the strongest possible shape. But knowing myself, it won't be long before I'm getting credit card alerts for my Cursor token spend.

[Drew Houston @drewhouston]

@arashf it sure was! ❤️🫡

[Drew Houston @drewhouston]

@levie @Dropbox game time 🤘

[Drew Houston @drewhouston]

I shared more with the Dropbox team this morning — congrats Ashraf! https://t.co/Ab80vrHyNk

[Drew Houston @drewhouston]

@andrewdfeldman @cerebras @Nasdaq Congrats Andrew!! 🥂

[Drew Houston @drewhouston]

@zeeg I'd just be guessing -- certainly local claw type use cases. But spark seems worse than apple silicon (high spec macbook or mac studio) at that price point given slow mem bw (273GB/sec vs ~800GB/sec with apple or 1.7TB/sec with a 5090 or RTX pro 6000)

[Drew Houston @drewhouston]

@kevinweil Congrats @kevinweil! 🤘

[Omar Khattab @lateinteraction]

Wow! Quite a detailed blog from @Dropbox on how they optimize Dropbox Dash relevance judge with @DSPyOSS.

Shout out to the Dropbox Dash team Eric Wang, Dmitriy Meyerzon, and @joshclemm !

[Drew Breunig @dbreunig]

Incredible case study that's readable yet filled with practical approaches and tactics:

✔ Use GEPA to get near-frontier performance at 1/100th the cost.
✔ Quickly evaluate new models, without having to rewrite prompts.
✔ Evolve & maintain your programs w/o full rewrites

[Drew Houston @drewhouston]

@__tinygrad__ congrats!

[Drew Houston @drewhouston]

@jxmnop Actually pretty much as described. GLM 4.7 runs at ~75-80 tok/sec (bs1, EAGLE3.) But not out of the box. To fit 355B params in 384GB I hacked up sglang to support MoE GEMMs in mxfp6, writing cutlass/triton kernels for blockscaling & sm120 support. "Quite trivial"