swalsh

8 ms·

91.
by swalsh 1y ago
> can't work well because of motion sickness. This is an overated problem. You play VR for a small amount of time then you adapt to it. You get your "VR Legs" as they say.
92.
by swalsh 1y ago
I LOVED VR gaming, but after playing the same 2 games for 10 years, it never really evolved. They stopped innovating and went all in on AR.
93.
by swalsh 1y ago
The people shipping these features are not the same people who are fixing reliability probably.
94.
by swalsh 1y ago
That models entire world is the corpus of human text. They don't have eyes or ears or hands. Their environment is text. So it would make sense if the environment contains human concerns it would adopt to human concerns.
95.
by swalsh 1y ago
To be fair, I'm seeing the demo video, and I still don't believe it's possible. This is sci-fi tech.
96.
by swalsh 1y ago
To me it depends on 2 factors. Hardware becomes more accessible, and the closed source offerings become more expensive. Right now it's difficult to get enough GPUs to do local inference at production scale, and 2 it's more expen
97.
by swalsh 1y ago
Oh I wonder if that applies to me? I've been using claude to do experiments with using SNN's for language models. Doubt anything will come of it... has mostly just been a fun learning experience, but it is technically a "c
98.
by swalsh 1y ago
While it's being generated, I'll spot check it, and after I test the code i'll peek in more detail at it. I review the code in much the same way I review code from a human dev. I almost never look closely at ALL lines. I&#
99.
by swalsh 1y ago
That's fine, please make it VERY CLEAR how much of my limit is left, and how much i've used.
100.
by swalsh 1y ago
I used about $300 worth of credits based on ccusage ($20 pro plan). It's pretty easy to hit the limit once you get going.
101.
by swalsh 1y ago
This happens when you prompt it poorly. If you want to avoid slop, the first step is to write an extensive BRD. Read it, understand it, make sure it has everything needed. Then write a solutions architecture document. Read it, understan
102.
Hierarchical Reasoning Model(twitter.com)
2 points by swalsh 1y ago|0 comments
103.
by swalsh 1y ago
I built something similar for my own workflow. Works okay. The hard part is as you scale, you end up with compounded false affirmatives. Model adds some fallback mechanism that makes it work, tests pass, etc. The nice part is you can as
104.
by swalsh 1y ago
The answer to "how many racks are they selling" is currently as much as they can manufacture, extended out a year.
105.
by swalsh 1y ago
Chatbot advertising has to be one of the most powerful forms of marketing yet. People are basically all the way through the sales pipeline when they land on your page.
106.
by swalsh 1y ago
Id guess the answer is gpt4o is an outdated model that's not as anchored in reality as newer models. It's pretty rare for me to see sonnet or even o3 just outright tell me plausible but wrong things.
107.
by swalsh 1y ago
You can just not post if your criticism is mean spirited.
108.
by swalsh 1y ago
The hard part of building an agent is training to model to use tools properly. Fortuantely Anthropic did the hard part for us.
109.
by swalsh 1y ago
I don't know if you understand the role the LLM is playing here. The mechanism used to execute the command is not the relevant thing. The LLM autonomously executing commands has intelligence, it's not just a shell script. If I
110.
by swalsh 1y ago
Elicitation is a big win. One of my favorite MCP servers is an SSH server I built, it allows me to basically automate 90% of the server tasks I need done. I handled authentication via a config file, but it's kind of a pain to manage
111.
by swalsh 1y ago
Sonnet 4 changed my mind on AI safety. It can do ALOT of work unattended, real work like configuring servers. If you give it a goal, and a set of tools, it will get the job done. But I got freaked out the first time I used it, since I di
112.
by swalsh 1y ago
I ran into this issue, I built my own bash and SSH MCP server. In my first iteration I did not quite trust Claude yet so I limited the commands it was allowed to run in Bash. But I gave it access to Python, so any time it ran into a limit
113.
by swalsh 1y ago
I read LLM generated code like I review a PR. I skim for anything that stands out as a common pitfall, and dig into the details of area I expect issues.
114.
by swalsh 1y ago
You can ask the llm to write code the way you think about it. I usually write a little spec document as my first prompt, and in there i'll add some rules on how I want code structured, seperated etc. If you use cursor, cursorrules can
115.
by swalsh 1y ago
> Tech workers should be organizing to prepare for the profit-taking moves management has in store for us I think you think this is going to help tech workers, but raising the cost of employing humans is only going to incentivize compani
116.
by swalsh 1y ago
This game feels like someone trying to RLHF some license plate model
117.
by swalsh 1y ago
I think the non-creative work is coming... but it's harder, needs more accuracy, and just generally takes more effort. But it's 100% coming. AI today can one shot with about 80% perfection. But for use cases that need to be hig
118.
by swalsh 1y ago
that's neat
119.
by swalsh 1y ago
A tired old debate devoid of facts. The chains I advocated for are proof of stake, and AI is a useful system worth the energy it uses. Should we not consume energy at all?
120.
by swalsh 1y ago
clearly, that is not what I said. What I said is there is a very big problem (closed ecosystem of incompatible credit systems) that can be solved using this technology that does payments really well.

More ›