🔧 AI Nachrichten Major AI platforms go down in unprecedented simultaneous outage(03.09.2026 um 17:34 Uhr)
🔧 AI Nachrichten ChatGPT, Claude, and Grok Down? Users Report Widespread Outages(03.09.2026 um 19:14 Uhr)
🔧 AI Nachrichten OpenAI Launches GPT-6 Astra, Says We May Have Entered the AGI Era(03.09.2026 um 22:08 Uhr)
🔧 AI Nachrichten Claude Comes to CarPlay as Fifth Major AI Chatbot App(05.09.2026 um 05:31 Uhr)
🔧 AI Nachrichten OpenAI’s GPT-6 Astra Is AGI, Says NVIDIA CEO Jensen Huang(07.09.2026 um 06:31 Uhr)
🔧 AI Nachrichten Blame AI companies for Mac mini and Mac Studio shortage(31.08.2026 um 10:32 Uhr)
🔧 AI Nachrichten Major AI platforms go down in unprecedented simultaneous outage(03.09.2026 um 17:34 Uhr)
🔧 AI Nachrichten ChatGPT, Claude, and Grok Down? Users Report Widespread Outages(03.09.2026 um 19:14 Uhr)
🔧 AI Nachrichten OpenAI Launches GPT-6 Astra, Says We May Have Entered the AGI Era(03.09.2026 um 22:08 Uhr)
🔧 AI Nachrichten Claude Comes to CarPlay as Fifth Major AI Chatbot App(05.09.2026 um 05:31 Uhr)
🔧 AI Nachrichten OpenAI’s GPT-6 Astra Is AGI, Says NVIDIA CEO Jensen Huang(07.09.2026 um 06:31 Uhr)
🔧 AI Nachrichten Blame AI companies for Mac mini and Mac Studio shortage(31.08.2026 um 10:32 Uhr)

🔧 Programmierung 🕛 kürzlich 11 Min Lesezeit
0

Take your benchmark to the people who can kill it

↗ Quelle (dev.to)
🗣️ Stimme:
📑 Inhaltsübersicht

Ten days ago I had a benchmark: reorder the expert weights inside a MoE model file by measured co-activation, and disk reads per token drop 2.23×. It was real, reproducible, and measured entirely on my own machine — which is to say, it was worth almost nothing yet.



Today the strongest number attached to this work is one I didn't produce: +32.3% decode throughput and −26.3% time-to-first-token on a 235B-parameter model running on a 48 GB MacBook, measured by someone I've never met, on an inference engine I'd never run, using my script unmodified, with the arms swapped as a page-cache control. Along the way, two of my three original pitches were refuted with data, my one pre-registered scaling prediction failed its own threshold, and both of my attempts to forecast a key parameter missed in opposite directions.



The refutations were the productive part. This post is about the process that made them cheap: five engines, a half-dozen strangers, one issue thread, and the discipline that emerged in it — because I think the process generalizes and the write-ups I usually see don't mention it.






The setup



The idea (project: , where @mabaeyens was shipping an expert-offload streamer for MLX, and measured his 744B engine's container — which already stores each expert as one contiguous ~19 MB read — at a residual +1.3%, below his own promotion gate. He posted it as a no-go, I agreed, and the ~+20% GLM-class number I'd floated earlier was retired. The coalescing ladder is now fully priced: nine-scattered → +14–32% depending on share; per-projection → +31% at deep offload; per-expert-contiguous container → approximately nothing, at device bandwidth ceiling. An optimization whose domain of inapplicability is measured and published is worth more than one with only success stories.






Nulls are product



The thread's graveyard, as of today: predictive prefetch (three independent nulls — the structural one, the properly-controlled A/B, the cross-architecture shuffle test), heat-pinning versus LRU, smarter-than-LRU eviction, and my original cross-layer framing (retracted). Each null is scoped — batch-1, decode, these architectures — and each one is a roadmap item somebody now doesn't have to build.



Then the ending wrote itself. Late in the thread, PhilipJohnBasile open-sourced his engine — and its README turned out to contain an independent replication of nearly the entire design theory, measured on a fifth engine before we'd ever spoken: router-lookahead prefetch reverted at −6.96%, smarter eviction dead at −26% with zero hit-rate change, static pinning dead, hotset training dead. His summary line: "the cache-policy frontier is closed: capacity vs reuse distance is the wall." We had converged on the same wall from four different codebases.



(That open-sourcing also resolved a citation mystery in one of the great small moments of the thread: a throughput figure everyone kept attributing to our measurements turned out to be an Italian code comment at line 6141 of the parent project his engine forks — measured on Linux in early July, a platform qualifier his fork's copy had dropped. Two of us had "verified" the wrong provenance because GitHub's code search silently doesn't index files that large. That's in the traps section now too.)






What I took away





  • Take claims to the people with the machines to kill them. Four engine authors falsified, scoped, or extended every major claim within days. No general audience does that. The niche thread was not the humble alternative to a launch; it was the higher-bandwidth channel.


  • Pre-registration is what makes being wrong cheap. Every number I committed to in advance either survived (and now carries weight) or died cleanly (and produced a law). The two forecasts I hedged on are the ones that cost follow-up rounds to untangle.


  • Public concession is the currency. The thread's author said it at the close, better than I can: "I learned more from the parts where I was wrong than the parts where I was right. Good thread."


  • Keep one versioned, red-line-able summary. It outperforms both the scrollback and any paper: it's current, it's editable, and every participant has standing to correct it.


  • Ship your nulls with your wins. The parity result on stock llama.cpp, posted up front, is why the right people engaged at all.



The thread is still open. Co-activation-aware ordering is now the named follow-up on the mlx-lm PR; the colibri community is building an expert atlas with the co-activation format from this work; there's a co-signed, pre-declared cross-engine experiment waiting on two routing traces. The consolidated summary is , and the repo is (the 744B endpoint, iliria), @pierre427 (production pager, the registered-band point), and the colibri crew — @JustVugg, @ZacharyZcR, @bokiko, @mohamedmastouri2000-boop — who are taking the co-activation work somewhere none of us planned.



For more information and other projects go to https://getnadir.com

Vollständiger Original-Bericht
Ausführliche Details, Code-Beispiele & Hersteller-Stellungnahme auf dev.to.
↗ Original-Artikel auf dev.to lesen
Wie bewertest du diesen Beitrag?
1 Klick Feedback
Teilen mit Netzwerk & Team:

Community-Analysen & Experten-Meinungen 0

Verfasse deine eigene Analyse, teile Workarounds oder diskutiere diesen Vorfall im Blog.
Noch keine Community-Analyse verfasst. Markiere einen Textabschnitt oder klicke oben auf Eigene Analyse verfassen“!
Community Pulse: Relevanz-Einschätzung
1 Klick Experten-Votum
🔴 Akute Relevanz 0%
🟡 In Evaluierung 0%
🟢 Keine Auswirkung 0%
Spannende Innovation 0%
Verwandte Story-Cluster & Quellen (Vektor-KI)
Port 8095 Engine
3 Quellen
GPT-6 Astra Release Today? OpenAI’s Next Major AI Model Is Almost Here
1 Quelle
Apple accuses OpenAI of destroying evidence as trade-secrets fight intensifies
1 Quelle
Major AI platforms go down in unprecedented simultaneous outage
Ähnliche Beiträge
🔍 Verwandte News

Auch interessante Nachrichten Take your benchmark to the people who can kill it

Thematisch verwandte Begriffe: Take, your, benchmark, people · 6 Treffer

Laden...

Videos werden geladen ...

Laden...

Beiträge werden geladen ...

Laden...

Videos werden geladen ...

Laden...

Beiträge werden geladen ...

Laden...

Videos werden geladen ...

Laden...

Beiträge werden geladen ...

Laden...

Videos werden geladen ...

Laden...

Beiträge werden geladen ...

Laden...

Videos werden geladen ...