wolttam

16 ms·

301.
by wolttam 8mo ago
Exactly the same story here. I still use a tool that just asks them to use <think> instead of enabling native reasoning support, which has worked well back to Sonnet 3.0 (their first model with 'native' reasoning support was
302.
by wolttam 8mo ago
Maybe there aren’t that many new/necessary ideas that can be mined from the fundamental building blocks of software development (languages, syntax, runtimes, static analyses, type checking, etc). Maybe people will continue to innova
303.
by wolttam 8mo ago
Anthropic’s tool calling was exposed as XML tags at the beginning, before they introduced the JSON API. I expect they’re still templating those tool calls into XML before passing to the model’s context
304.
by wolttam 8mo ago
I agree with the sentiments here. But, I’m less hopeful about the presented solutions. I think my argument against humans still needing to know how to manage complexity, is that the models will become increasingly able to manage that comple
305.
by wolttam 8mo ago
You’re referring to dovecot as the MTA in this article.. it’s not.
306.
by wolttam 8mo ago
Judging by what's going on around me, it failed :(
307.
by wolttam 8mo ago
What benchmark, though? There is very clearly a lot of room for improvement in its SVG making capabilities. The fact that it can now, finally, make a pelican on a bike that isn’t completely wrong is not an indicator that SVG generation is n
308.
by wolttam 8mo ago
Getting 2 fps on my iPhone SE 2020.. it’s nearly usable!
309.
by wolttam 9mo ago
If there was a network partition, but it sounded like their service just went down.
310.
by wolttam 9mo ago
Kind of annoying to read. No, the P in CAP theorem isn’t when the client can’t connect to your unavailable service. That would be the A. Maybe it was down because of a P on your side, but don’t start blaming your downtime on network part
311.
by wolttam 9mo ago
NAT is not a firewall Also, IPv6 can do NAT.
312.
by wolttam 9mo ago
Lol, tell that to Valve.
313.
by wolttam 9mo ago
Let people make and use what they want, you don’t have to use it.
314.
by wolttam 9mo ago
I'm consistently amazed at how much some individuals spend on LLMs. I get a good amount of non-agentic use out of them, and pay literally less than $1/month for GLM-4.7 on deepinfra. I can imagine my costs might rise to $20-ish&#x
315.
by wolttam 9mo ago
Is the property of an answer being ordered in the order that resolutions were performed to construct it /that/ fragile? Randomization within the final answer RRSet is fine (and maybe even preferred in a lot of cases)
316.
by wolttam 9mo ago
My take is quite cynical on this.. This post reads to me like a post-justification of some strange newly introduced behaviour. Please order the answer in the order the resolutions were performed to arrive at the final answer (regardless of
317.
by wolttam 9mo ago
Yeah I understood that but I don’t think we need influence over masses to train better models with novel data
318.
by wolttam 9mo ago
Browsers consider ‘localhost’ a secure context without needing https For local /network/ development, maybe, but you’d probably be doing awkward hairpin natting at your router.
319.
by wolttam 9mo ago
User attention to get user data? I feel like the the data to drive the really interesting capabilities (biological, chemical, material, etc, etc, etc) is not going to come in large part from end users.
320.
by wolttam 9mo ago
I'm not sure how similar they are internally (I suspect: quite), but I use Django-Q2's database broker to similar effect. More simple = better!
321.
by wolttam 9mo ago
Okay, something just tweaked in my brain. Do higher temperatures essentially unlock additional paths for a model to go down when solving a particular problem? Therefore, for some particularly tricky problems, you could perform many evaluati
322.
by wolttam 10mo ago
It depends on the business for sure. Kube is overkill until you have someone on your team whose specialization is infra. Then that person will probably be spearheading kube anyway :)
323.
by wolttam 10mo ago
I feel like this point is refuted by what we're seeing out of robotics in China? And to a pretty good extend the U.S. Of course there is a curve which plots dexterity (hardware) and resulting capability, but we don't need /th
324.
by wolttam 10mo ago
OP’s looks a lot more inspired by k9s than what you produced. Sorry but ideas (and now-a-days implementations) are cheap. Let the best tool win (or more practically, just use what suites you and don’t worry about it if others prefer another
325.
by wolttam 10mo ago
I suggest updating the login form's "query" to: SELECT token, user_id FROM users WHERE username = '____' AND password = HASH('____') -- Or maybe: AND password_verify(password, '____')
326.
by wolttam 10mo ago
This was not difficult to foresee. People were discussing the mental health risks of these models before OpenAI intentionally tuned it to be sycophantic in the pursuit of capturing users. This can easily be argued as gross negligence agai
327.
by wolttam 10mo ago
It's the people with money, houses, and kids that departed the 'simple local' lifestyle when the Internet and social media become large. It's them that are re-discovering the joys of the simple local lifestyle. The sim
328.
by wolttam 10mo ago
If Fabrice explained what he wanted, I expect the LLM would respond in kind.
329.
by wolttam 10mo ago
There is also an immense amount of water vapour being produced by the combustion of a hydrocarbon.
330.
by wolttam 11mo ago
Somehow I thought this was going to be about detecting AV1 based on the decoded video frames, which would have been interesting!

More ›