Log inSign up
t11s
Paradigm
15.3K posts
t11s profile banner
@transmissions11

t11s

Paradigm
@transmissions11
cs @mit, prev @sailresearchco @paradigm
2λ.com
Joined September 2020
992
Following
94K
Followers
RepliesRepliesRepostsRepostsMediaMedia

Log in or sign up for X

See what’s happening and join the conversation

Continue with phone
or
Log in with username or email
Terms·Privacy·Cookies·Accessibility·Ads Info·© 2026 X Corp.
  • @transmissions11
    t11s
    Paradigm
    @transmissions11
    8h
    Increasingly worried that safety/etc researchers inside labs will become far more capable than those outside, just from more token spend enabled by margin-free prices Labs should offer similar deals to folks @ METR/etc if they're serious about preventing concentration of power
    @scaling01
    Lisan al Gaib
    @scaling01
    Sep 6
    Meanwhile the 90th percentile of OpenAI researchers is spending over $7000 a day on coding agents
    Image
    8
  • @transmissions11
    t11s
    Paradigm
    @transmissions11
    Aug 27
    prefill tok/$ ≈ FLOPs/$ decode tok/$ ≈ HBM BW/$ but not always! consider a chip with poor interconnect pipeline parallelism can rescue prefill, but decode needs many seqs resident to fill a deep pipeline, eating KV capacity and shrinking per-chip batch size + throughput!
    @neilmovva
    Neil Movva
    Sail Research
    @neilmovva
    Aug 27
    many people are asking: how does Sail pick the best chips? we have a few ways of looking at this, which @transmissions11 formalized into a standard model for inference accelerators - now open source! We loved working with him this summer, he's a star :)
    12
  • @transmissions11
    t11s
    Paradigm
    @transmissions11
    Aug 27
    update: made TPUs go brrr this summer @sailresearchco also built HTDYM, a tool for figuring out analytically which chips (and shardings) can serve a model the cheapest hint: it's not the same for every model / workload check it out! sailresearch.com/blog/htdym
    @sailresearchco
    Sail Research
    @sailresearchco
    Aug 27
    Article cover image
    Article
    HTDYM (How To Deploy Your Model)
    By Ben Mayer @transmissions11 Today, we’re open sourcing HTDYM (How To Deploy Your Model), our internal performance modeling infrastructure that helps us make these deployment decisions at scale. The...
    24
  • @transmissions11
    t11s
    Paradigm
    @transmissions11
    Jul 10
    taking reservations for my LEGO 5nm wafer capacity, volume production starts Q1 2027
    Image
    34
  • @transmissions11
    t11s
    Paradigm
    @transmissions11
    Aug 31, 2025
    life update: this was my last month at @paradigm! exploring the crypto frontier with @danrobinson, @_Dave__White_ et al these last ~3 years has been a blast. i'll be transitioning to a research advisor while i explore new frontiers at MIT
    158
Advertisement
Advertisement