htrp

13 ms·

361.
by htrp 10mo ago
reminder that according to Facebook's own analysis 10% of their 2024 revenue comes from scams and banned products
362.
by htrp 10mo ago
All tech problems are actually people problems. once the Csuite builds their own dashboards, they quickly decide what they actually need versus what is a nice to have.
363.
by htrp 10mo ago
the initial rlhf training evaluation was done by kenyans specifically
364.
by htrp 10mo ago
it's easy to scan for publicly known services, really difficult to understand if a random string that says key somewhere is actually a random internal api key
365.
by htrp 10mo ago
much harder to train longer context inputs
366.
by htrp 10mo ago
Everyone trains on queries from other models, it's called distillation
367.
by htrp 11mo ago
what's wrong with the current ide tools?
368.
by htrp 11mo ago
I feel like that's how you get Microsoft where each division has a gun pointed at the other division
369.
by htrp 11mo ago
that sounds like a pm problem
370.
by htrp 11mo ago
depressingly enough, things that work on small scale architectures often don't work at larger scales
371.
by htrp 11mo ago
much easier to raise money as a frontier lab
372.
by htrp 11mo ago
Panamax ship is 5000 teu (twenty foot shipping container equivalent) I think you get about 4 MWh per TEU ( based on my 12V 100Ah battery) so about 20 GWh
373.
by htrp 11mo ago
>the original Microsoft business model From 1981 >Microsoft, which needed an operating system for the IBM Personal Computer,[9][10] hired Tim Paterson in May 1981 and bought 86-DOS 1.10 for US$25,000 that July
374.
by htrp 11mo ago
For people who are unclear what Palantir actually does Palantir is a tech platform that consumes data from their clients in return for providing high level data-driven insights. They assign FDEs (or consultants) to really learn the details
375.
by htrp 11mo ago
> IMO the only remaining unanswered question for the Starship program is the reusability of the heat shield. There's no reason to believe any other part of it can't work. Is there a tldr someone put together here ?
376.
by htrp 11mo ago
it wasn't economical to deploy but i expect it wasn't wasted, expect the openai team to pick that back up at some point
377.
by htrp 11mo ago
spoiler alert, they don't work without a lot of custom code
378.
by htrp 11mo ago
Trinity Nano Preview: 6B parameter MoE (1B active, ~800M non-embedding), 56 layers, 128 experts with 8 active per token Trinity Mini: 26B parameter MoE (3B active), fully post-trained reasoning model They did pretraining on their own and ar
379.
by htrp 11mo ago
Anyone remember Soylent?
380.
by htrp 11mo ago
isn't antigravity the windsurf team?
381.
by htrp 11mo ago
The problem is asking for user preference leads to sycophantic responses
382.
by htrp 11mo ago
what is the ballpark vram / gpu requirement to run this ?
383.
by htrp 11mo ago
You can if you are Meta and are willing to litigate the hell out of it.
384.
by htrp 11mo ago
So what's the difference between their announcement in april and now?
385.
by htrp 11mo ago
> Google interview? did he do something other than that podcast?
386.
Warner Music Settles Lawsuit with AI Music Startup Suno(bloomberg.com)
2 points by htrp 11mo ago|1 comments
387.
by htrp 11mo ago
og web3- semantic web
388.
by htrp 11mo ago
extending the example, for PE, the mortgage is actually in the name of the house and you can take out additional loans in the name of the house to pay yourself the down payment that you used to purchase the house (while simultaneously strip
389.
by htrp 11mo ago
Google AI products are the new chat product
390.
by htrp 11mo ago
> The approach relies on an extreme decomposition of a task into subtasks, each of which can be tackled by focused microagents. The high level of modularity resulting from the decomposition allows error correction to be applied at each s

More ›