vessenes

26 ms·

481.
by vessenes 8mo ago
I'll bite. I think their conversation (voice) model is more fluid than competitors. It's also very good at hitting up twitter for realtime information, and was that way before the current tool use models got fully up and running.
482.
by vessenes 8mo ago
There's some texture here. Elon's enriched pretty much everybody who's ever worked for and invested with him. He makes money for people throughout his orgs. Many ex-employees have said to me: "incredible opportunity, mad
483.
by vessenes 8mo ago
Email summarization, calendar notifications, briefing documents, .. the list goes on. Think of anything knowledge-based that's moderately repetitive you don't want to do, and just ask
484.
by vessenes 8mo ago
Absolutely no relation to these guys, but I have a hobby knitwear line and a longstanding fascination with Shima Seki. The SS machines are unique, as far as I know. They are also SUPER hard to work with -- the software is opaque. The design
485.
by vessenes 8mo ago
I have a small sweater line I’m looking at doing in china right now but I have a long lasting fascination with the shima seki machines. If you were doing a short fashion run would TI be appropriate? How does one get their preferred yarn ove
486.
by vessenes 8mo ago
I could not get my Mac to successfully do anything with the script from their GitHub; set device to mps, downloaded llama for the first time in a year, and it just .. hangs. I presume this is sortable, but I'm not sure I care enough. A
487.
by vessenes 8mo ago
It loads after those errors. Tap space and talk to it.
488.
by vessenes 8mo ago
Just tried it. really cool, and a fun tech demo with rcli. I filed a bug report; not everything is loading properly when installed via homebrew. Quick request: unsloth quants; bit per bit usually better. Or more generally UI for huggingface
489.
by vessenes 8mo ago
Thank you for real data. Please moderate the use of the word profitable talking to engineers! We get the same circle jerk over and over here. Profit implies a GAAP accrual of some sort. On any accrual schedule tied to reality, the companies
490.
by vessenes 8mo ago
Partner in a fund that wrote a small check into this — I have no private knowledge of the deal - while I agree that one’s opinion on auto regressive models doesn’t matter, I think the fact of whether or not the auto regressive models wo
491.
by vessenes 8mo ago
And on the DC side they will be building to a power and heat budget. If Vera Rubin changes the power density per rack equation that may have some impact. But thinking rationally if the flops per kw-sq ft are lower than Blackwell, no problem
492.
by vessenes 8mo ago
I'm curious what firm they chose for representation.. I guess I'd pick either Susman Godfrey (Fox dominion win) or whomever represented SpaceX when they were trying to get their first NASA contracts.
493.
Anthropic Sues DoD(wired.com)
5 points by vessenes 8mo ago|2 comments
494.
by vessenes 8mo ago
As someone who owned a dreamcast and stayed in line all night for a PS2, I can say that the DVD player functionality was HUGE. It was many, many people's first DVD player, and priced so that the game console was nearly 'free'
495.
by vessenes 8mo ago
Heading down the links of this blog ends up at https://github.com/gfrmin/credence , which claims to be an agentic harness that keeps track of usefulness of tools separately and beats LangChain at a benchmark. LangChain…
496.
by vessenes 8mo ago
Thanks. Similarly no comment there. Weird.
497.
by vessenes 8mo ago
Good thoughts. Welcome to the discourse. A couple things to put out there - first, the US has a fairly strong rule of law that the government cannot compel speech -- essentially, while speech can be blocked/stopped, it's a hard
498.
by vessenes 8mo ago
I know that you vibecoded this and I hope that you had fun building it. It does not have any commercial value in my opinion, at least under the use cases you show. Bascially you built out a tool that ‘fixes’ a problem nobody has - you’ve wr
499.
Ask HN: Why Is Phil Wang / Lucidrains Off GitHub?
4 points by vessenes 8mo ago|4 comments
500.
by vessenes 8mo ago
I have mine reading yours right now. Unfortunately(?) I mentioned LeCun to it, and it says it's adding a "causal world-state mixer" to nanograd; not sure how this will work out, but it wasn't nervous to do it. Gpt 5.4 xh
501.
by vessenes 8mo ago
On the skill side, personalities could be fun: "You are Yann Lecun's last PhD candidate, and he hates you and you hate JEPA. You are determined to prove that a non-world model can reach AGI. In order to get your PhD you have to be
502.
by vessenes 8mo ago
I asked Gemini for a literature search and it came back with this: References Chen, R., Arditi, A., Sleight, H., Evans, O., & Lindsey, J. (2025). Persona Vectors: Monitoring and Controlling Character Traits in Language Models. arXiv. h
503.
by vessenes 8mo ago
Introspection: all heard. As a practical matter, you can rl or prompt inject information about the model into context and most major models do this, not least I expect because they’d like to be able to complain when that output is taken for
504.
by vessenes 8mo ago
Anthropic's mechanistic interpretation group disagrees with you - they see similar activations for 'hallucinations' and 'known lies' in their analyses. The paper is pretty interesting actually. So, you're wrong
505.
by vessenes 8mo ago
You can learn a lot from a model when you ask about its sizing, although not necessarily anything about the sizing. For instance, you can learn how much introspection has been trained in during RL, and you can also learn (sometimes) if outp
506.
by vessenes 8mo ago
I know. If you have an old iPod, try it. I stand by my review; I think it's probably 3x as much latency.
507.
by vessenes 8mo ago
Just to parse this out, I think y'all are correct at the reading - that people didn't want to give up their custom workflows to fit in with the times. In my defense, I was up early. I stand by my review - her entire response is ex
508.
by vessenes 8mo ago
Like any doubling rule, the buck has to stop somewhere. Higher energy usage + smaller geometry means much more exotic analog physics to worry about in chips. I’m not a silicon engineer by any means but I’d expect 10Ghz cycles will be optica
509.
by vessenes 8mo ago
Some real gold from SuburbanWhiteChick in the comments: Fifth. Computerization has not improved standards; it has merely homogenized them. When humans do work, even soul-killing work, they either get bored and get out or they start to s
510.
by vessenes 8mo ago
Please enlighten me.

More ›