Favikon Index
it
← Top 100 IA
HHHorace He, Engineer, Thinking Machines

Top 100 IA · Stati Uniti

Horace He

Thinking Machines Stati Uniti

Punteggio
6575+1 oggi
Posizione
103/198 ↓24
in 90 giorni
−225−3,1%

Horace è un ingegnere del software specializzato in machine learning e compilatori, con un focus sui sistemi di intelligenza artificiale e sull'interazione uomo-AI. Condivide approfondimenti tecnici, confronti sulle prestazioni e contenuti educativi riguardanti modelli di ML, ottimizzazione hardware e soluzioni di codifica. I suoi post spesso includono benchmark, risultati di ricerca e meme legati all'intelligenza artificiale e all'ingegneria del software.

CondividiXWhatsAppLinkedInFacebook

Dati al 25 ott 2026.

I numeri grezzi · ultimi 90 giorni

Visualizzazioni mediane
27,9K
per post
Interazioni mediane
243
per post
Post
24
in 90 giorni
Follower
56K
tutti i social nel punteggio

Cosa determina il punteggio

Percentili 0 – 100 rispetto a tutti i creator monitorati.

  • Portata · 30%81

    Visualizzazioni medie per post su 90 giorni

  • Engagement · 30%60

    Reazioni medie per post

  • Audience · 30%59

    Follower totali, rapportati alla popolazione

  • Costanza · 10%98

    Ritmo di pubblicazione (25 post / 90 g = 100)

Social

56K follower in totale

SocialFollowerPunteggio
X@chhillee54,9K7984t2
LinkedIn@horacehe1,1K35t10

Ultimi 30 giorni

I migliori post di He

  1. X · l’altro ieri#1
    I'm doing DP. I'm doing TP. I'm doing SP. I'm doing PP. I'm doing ZP. I'm doing FSDP. I'm doing 1F1B. I'm doing CP. I'm doing ETP. I'm doing DPA. I'm doing VPP. I'm doing P/D. I'm doing ZBPP. I'm doing DPA. I'm doing …
    HHHorace He, Engineer, Thinking MachinesHorace He

    41,8K visualizzazioni713 interazioni

  2. There's an interesting distortion in inference pricing that appears to be driven by openrouter's routing. Namely, ridiculously cheap prefill and very expensive decode. For GLM 5.3, https://t.co/aLgdvVsDHp (highest vol…
    X · 22 gg fa#2
    HHHorace He, Engineer, Thinking MachinesHorace He

    There's an interesting distortion in inference pricing that appears to be driven by openrouter's routing. Namely, ridiculously cheap prefill and very expensive decode. For GLM 5.3, https://t.co/aLgdvVsDHp (highest vol…

    57,5K visualizzazioni686 interazioni

  3. X · 8 gg fa#3
    Your model behaves differently during training and inference? Boy do I have a solution for you :^) https://t.co/60WkGyfwfD
    HHHorace He, Engineer, Thinking MachinesHorace He

    30,2K visualizzazioni524 interazioni

  4. X · 27 gg fa#4
    I find that surprisingly few people have *really* thought about why PD disagg is a good idea. Even fewer people have *really* thought about how PD disagg compares to mixed chunk prefill (as Matthew describes here).
    HHHorace He, Engineer, Thinking MachinesHorace He

    113K visualizzazioni468 interazioni

  5. X · 12 gg fa#5
    As @johnschulman2 has put it, there are many ways we can "spend compute to save our customers compute". Here's one of them - a variant of muon that works far better for lora fine-tuning :)
    HHHorace He, Engineer, Thinking MachinesHorace He

    25,7K visualizzazioni329 interazioni

  6. X · 30 gg fa#6
    I think this is kind of an interesting idea, and in some sense Tinker is already a way to "sell" our pretraining stack (although tinker admittedly doesn't use the super large scale features), but I'm not really sure w…
    HHHorace He, Engineer, Thinking MachinesHorace He

    23,3K visualizzazioni248 interazioni

Faccia a faccia

Confronta He con…