Log inSign up
Log inSign up
Kevin Madura
2,489 posts
Kevin Madura profile banner
@kmad

Kevin Madura

@kmad
💻 | kmad.ai
DC
Joined May 2008
1,238 Following
1,805 Followers
RepliesRepliesRepostsRepostsMediaMediaArticlesArticles

Kevin Madura’s posts

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·US TIDA·Ads Info·© 2026 X Corp.
  • @kmad
    Kevin Madura
    @kmad
    58m
    Well written
    @john_ssuh
    John Suh
    @john_ssuh
    2h
    Agent platform of 2027. This is the end game. This is the blueprint for the one and only application layer software that will exist. Every other software will effectively become databases for this layer, the everything agentic operating system for the million agent swarm to
  • @kmad
    Kevin Madura
    @kmad
    4h
    capitalism is a beautiful thing
    @thdxr
    dax
    Anomaly
    @thdxr
    7h
    we are popping the pricing bubble today starting with deepseek flash and glm 5.3 we are pricing 60% lower than fireworks/baseten while doing more volume than them all hosted on our own GPUs, there’s no reason to buy from anyone else we’re #1 now, good luck everyone else
    Quote
    @opencode
    OpenCode
    Anomaly
    @opencode
    7h
    GLM 5.3 and DeepSeek V4.1 Flash are now 60% off in OpenCode
  • @kmad
    Kevin Madura
    @kmad
    8h
    We’re lucky they’re even self reporting progress in cryptanalysis … imagine what isn’t published elsewhere
    @joy
    joy larkin
    @joy
    9h
    Replying to @kyleichan
    MiniMax released an interesting related blog post about cryptanalysis yesterday... minimax.io/blog/practical…
  • @kmad
    Kevin Madura
    @kmad
    11h
    Highly recommend reading this paper, it’s very approachable and the results are compelling - decomposing tasks such that they are in-distribution for sub-LM calls *works*. We just need to train models to do that decomposition effectively
    @himanshutwtxs
    hims
    @himanshutwtxs
    Nov 20
    TLDR of this awesome read: 1> Most harnesses keep appending the whole history into one prompt, so the model has to cope with contexts it never trained on 2> Harnesses like RLMs instead keep the context outside the prompt, write code over it, and hand smaller pieces to
  • @kmad
    Kevin Madura
    @kmad
    Nov 20
    ANT falling 27% after Claude solved the Reimann Hypothesis is not exactly what you’d expect