52 karma · joined April 29, 2025
> We cannot ignore individual benefits in contributing, but that is not everything.
Individual benefits are the very foundation of most collaborative efforts. Pharmaceutical companies get payouts via patents for researching life-saving medicine. Even with money taken out of the equation, many academics slave for years over their research in hope of recognition either via Tenure or Nobel Prize. Ignoring the fact that people want credit for their work is dangerous and irresponsible.
Forking projects that are near-complete or have already undergone major refactors just to snipe them last-minute with vibe coded efforts is going to kill collaboration. I'm never building another project in the open again because that is precisely what I'm aiming to avoid.
Some of the government bans have entered into the territory of banning the open weight model itself. This is like saying we must ban intelligence services or the military from using public weather reports in China for any operations. Public data is used all the time for these things, and an open weight model like DeepSeek should be treated with some level of care (e.g. putting it behind a firewall, having people study the model outputs, etc.) like you would for any foreign source of data, but that's all it is. data.
The hugging face hack was not a model gone rogue, it was intentional security malpractice. An open weight model cannot re-arrange electrons nor can it magically break a firewall that's setup properly.
Then why are the performance graphs the headliner topic in the article? The WASM sandbox is barely a footnote.
All you need is the source code for your existing project and the claude code/codex/opencode cli and boom, you are ready to ship. Most people doing the impressive stuff with AI are just using it, not building harnesses or "agentic" tooling.
I'm not a mathematician but I find this framing to be rather non-pragmatic and frankly, pointed.
In the process of Leonhard Euler's "rent-seeking" so he could "play with" bridge crossings he invented a completely new branch of mathematics that is the basis for things like Google Maps today. The process of working with these problems is just as important if not more, than just solving them.
I'm not a Musk fanboy by any stretch of the imagination but this was clearly done as a marketing stunt to drum up hype about the street legal roadster. I didn't even watch the announcement but now everyone in my friend circle is talking about it because of the rocket engines. That's just called effective marketing.
Tailscale pro pricing is $18 per user per month. Assuming you had one tech per store even, and 500 stores, that is a $9000 per month expense. Ok, not as bad as refrigerant and electric bill costs, but not great. So let's say instead of having a dedicated IT tech per store, we have 5 techs who work from HQ managing the infra for 500 stores.
$90. $90 per-month is all it would've taken to take these fridges off the public internet. If your margins are so razor thin that a monthly $90 bill would threaten profits, I question whether your business is even viable.
Reality is way more nuanced. Something can have high demand AND be on the precipice of financial catastrophe. See: Spirit Airlines.
Quite convenient how the same people trying to run roughshod over the others to get these bubble mania mausoleums built NEVER live near them but are HAPPY to make others who can't afford a second home live next to them.
SO tried to optimize the Q/A by eagerly rejecting repeat questions, without considering what the consequences of that were. If you’re in a classroom / lecture hall, your teacher / professor isn’t going to chide you for asking an “obvious” or “repeat” question (bad ones might). Imagine if every time you wanted to talk to a TA or professor you first needed to lookup previously asked questions in a database and then, iff your question hasn’t been asked before, you may ask the professor or TA after hours. Programmers might call this system “efficient” but most people would be pretty unhappy with it.
My only complaints are:
1) Memory usage is a bit high. We went with the AWS instance they recommended in the docs and even that went over our provisioned memory. It's not much but I think it could be improved
2) You need to buy their enterprise plan if what you're storing is remotely sensitive like health data, PII, etc. Any row level security or credential features are locked behind that license. Our use case isn't that sensitive so we can get away with putting it in a VPN and password protecting it, but if you need DB-level security the FOSS license is severely behind Postgres in terms of features.
Other than that, it's never gone down, it's very, very fast and comes with it's own webui for querying your data. We migrated from AWS Timestream and couldn't be happier with the switch.
What changes have LLMs (Not AI, not machine learning in general, I'm not going to waste time discussing the definition), LLMs made in the past 4 years that indicate anything close to the above? Solving a whiteboard math problem?