News for hackers. Stuff that matters.
Re:We Must Pace the Frontier by MisterTea on Saturday September 12, 2026 @07:06PM (#49678149)
I feel like all the closed model providers are milking it as they likely know open models on local hardware will one day eat their lunch. We all know it's not a matter of if but when. The company goes bankrupt, the hardware and property sold off, banks holding the bag.
Re:We Must Pace the Frontier by MichaelZuo on Saturday September 12, 2026 @07:38PM (#49678387)
Yeah avoiding all mention of the huge financial incentives that may push for “pacing the frontier” makes it seem like the opposite of a credibility boost for these firms.

It seems damaging since most folks (who lack insider knowledge) will naturally wonder if it’s due to plateauing performance per $ or some other non “alignment” reason.

Re:We Must Pace the Frontier by pvab3 on Saturday September 12, 2026 @08:39PM (#49678731)
The only way out is to develop a model vastly more powerful and capable that we have now. The market believes theres a good chance of that, although I've never understood why its truly winner-take-all
Re:We Must Pace the Frontier by taneq on Sunday September 13, 2026 @12:33AM (#49680048)
Because in the event that someone does build a strongly superhuman AI, no-one else will get a chance?
Re:We Must Pace the Frontier by kristofferR on Saturday September 12, 2026 @08:47PM (#49678782)
Cloud models will always have massive benefits of scale.

Caching is the simplest one to understand, cloud providers often reach a 90% cache hit rate, so hosting the same request locally on the exact same model on the same hardware is often way less efficient than on the cloud where a group of users generates a healthy cache.

Re:We Must Pace the Frontier by nixon_why69 on Saturday September 12, 2026 @09:56PM (#49679158)
KV cache is per conversation, I'm getting 100% hit rate on my single tenant local set up.

The benefits of scale are on the token generation side, you can batch rounds and generate tokens for multiple conversations per pass instead of just one token per pass.