efficax

6 ms·

61.
by efficax 2mo ago
all the effort in this space is going into multiuser, datacenter workflows for high throughput inference. running in resource constrained environments is not where the money is. but it will be, especially if we look forward to a world where
62.
by efficax 2mo ago
VCs tend to be a toxic bunch. Hard to avoid that in our business
63.
by efficax 2mo ago
Is this supposed to mean that 100k people live in each slice of the pie? Counting up slices and comparing to my metro statistical area population, that about works out.
64.
by efficax 2mo ago
Social security taxes explicitly pay for the current generation of retirees. it's not a pyramid scheme (how would that even make any sense?). The "trust fund" is running out of money but that could easily be solved if we we
65.
by efficax 2mo ago
I don't know if the masses can quite afford the 500k in GPUs you need to run this
66.
by efficax 2mo ago
in 2023 i wouldve said gpt could pass the turing test. today i could figure out it was an llm in a few turns no problem. llms cannot pass the turing test now that we’re accustomed to them
67.
by efficax 2mo ago
they are definitely not agi as it was ever defined. they’re only a bit more capable than they were a year ago. they crossed over from interesting crap to useful tool recently but really only for software
68.
by efficax 2mo ago
I feel like i've seen less hype about "the next model will be agi". GPT-6 is supposed to be coming this summer, and nobody is expecting AGI now. Not sure how they're going to keep the hype cycle going
69.
by efficax 3mo ago
will there ever be a way to fully audit american models? hell, at least they're releasing the weights for these so they actually could be audited!
70.
by efficax 3mo ago
there are lots of cloud suppliers besides whichever big name cloud this is, you can even find other cloud providers in seattle
71.
by efficax 3mo ago
so are buybacks. you choose to sign the contract. there's no way they didn't have an escape clause, although likely it meant not using the cloud provider anymore
72.
by efficax 3mo ago
how would you do that? you can copyright words, but you can't copyright an idea (you can patent some ideas, but not all of them)
73.
by efficax 3mo ago
they raised a ton of money but they spend it almost as fast as they can raise it. in any case they need real revenue if they want to ipo
74.
by efficax 3mo ago
I use a gemma4 model locally to extract content from messages to a personal agent I'm building for its memory graph (to break the message up into the topic, source (assistant or owner), facts, entities, etc. in the message content (all
75.
by efficax 3mo ago
for many purposes they're good enough now. If I had an opus 4.8 class model on a box next to me that could produce tokens at rates like 5000/s, i don't know if i'd need a new one for a long time. I think we might be unde
76.
by efficax 3mo ago
the distinction may be between using the coding agents vs using models for products. for example where i work we're talking about dropping opus for a chinese model for the in-app agent (which is very expensive to run)
77.
by efficax 3mo ago
Wrong adam. The adam who cofounded chef has nothing to do with opencode
78.
by efficax 3mo ago
If AGI comes to exist it won't be "priced" at all, since the lab that creates it will either quickly be seized by the gov't for national security, or they will become the most powerful organization in the world and have
79.
by efficax 3mo ago
only way you could win the market would be to own all the gpus
80.
by efficax 3mo ago
hey now there's plenty of ai fanatics in zip codes starting with 100 and 11
81.
by efficax 3mo ago
can you actually justify this claim. many of these technological advances were met with genuine public fanfare
82.
by efficax 3mo ago
if deepseek cost twice as much to train it would prove the same thing: the american companies have no monopoly on state of the art llms, and commoditization is happening
83.
by efficax 3mo ago
cost has nothing to do with why deepseek was disruptive, the fact that it means there is zero moat around anthropic or openai is what's disruptive about it. it means in the mid-term LLMs will be commoditized and customers will flock to
84.
by efficax 3mo ago
i pirated a ton but i also ripped all my cds and all my friends cds (and their parents cds). i took my macbook around everywhere and ripped every cd in could find
85.
by efficax 3mo ago
twist: codex also wrote the code that placed the winners list in a hidden element
86.
by efficax 3mo ago
the model does not make the architecture decisions, you make the architecture decisions. you define what success is, including performance characteristics. fuzzing and property based testing ensure correctness. what does it matter what the
87.
by efficax 3mo ago
But that's just more data. Create the error conditions in your test harness and exercise them. The machine will tell you what happened. Then correct the machine.
88.
by efficax 3mo ago
You build your outcome conditions and specifications in code as well. And you iterate on them too. I'm not talking about waterfall development. It's something else. You can spend a day just prototyping, spiking, experimenting, and
89.
by efficax 3mo ago
Code is not literature. It's not poetry. It does not express the human spirit. Code is a machine made of symbols. Engineers do care about the beauty of their machine designs, but in the end what matters is whether the machine performs
90.
by efficax 3mo ago
writing is a form of thinking, but there are others

More ›