alansaber

7 ms·

181.
by alansaber 2mo ago
Token reselling stops working when providers literally won't work with you
182.
by alansaber 2mo ago
When it comes to food contamination, even an infinitesimal quantity is noteworthy. Tbh many posts on this board are relatively low quality (and/or AI generated) but still spark interesting discussion.
183.
by alansaber 2mo ago
So what's the solution again? Everything as a web app?
184.
by alansaber 2mo ago
Github down or vibe-coded url. Call it.
185.
by alansaber 2mo ago
Ah yes no way global unification could go wrong. Send in the thunder warriors.
186.
by alansaber 2mo ago
Do they especially train for scientific performance?
187.
by alansaber 2mo ago
I love it when people handwave tech developments of a truly gargantuan scale
188.
by alansaber 2mo ago
If people can spam low-effort PRs they will. Unverified vibes is you closing your eyes to a very real issue.
189.
by alansaber 2mo ago
Vibe slop or put on your C.V "contributor to <wellknownproject>"? In terms of social currency B wins easily, unless you can guerilla market as well as Openclaw.
190.
by alansaber 2mo ago
Furnish makes more sense.
191.
by alansaber 2mo ago
The fit is theoretically good (nvidia is model agnostic) but I can see then pushing the ecosystem increasingly towards cuda and some custom nvidia software packages for inference etc etc
192.
by alansaber 2mo ago
Sounds like it was just too much money for the stakeholders to refuse
193.
by alansaber 2mo ago
Great for Nvidia. Bad for the rest of us. Oh well, so much for hugging face.
194.
by alansaber 2mo ago
We're all joking around here. But 5 decades down the line, with increasing cognitive offload to AI, is a slow erosion leading to this end state so crazy?
195.
by alansaber 2mo ago
Is the article slightly naive? Sure. It describes a relatively uncomplicated multi-agent system (which is good). But frontier model handing off a DAG rather than a raw markdown to a cheap model seems like a decent idea. Ultimately, this is
196.
by alansaber 2mo ago
Global international pressure seems to be the only solution to nautiucal problems in particular
197.
by alansaber 2mo ago
...mankind invented convenience?
198.
by alansaber 2mo ago
I have built systems using all of these approaches (all in tandem). For the most part, the juice is not worth the squeeze (in building a highly optimised corpus-specific information retrieval strategy) outside of a very few fringe cases. Th
199.
by alansaber 2mo ago
Isn't this just true for 90% of professional service industries?
200.
by alansaber 2mo ago
State consolidates power or a few megacorps. Unfortunately the middle ground (approximate balance of power) is always tenuous
201.
by alansaber 2mo ago
Yes yes it is key that we vibe code critical infrastructure. I can't wait to see a claude-ism in my next BIOS update.
202.
by alansaber 2mo ago
Quick, release a new claude product then depricate it after 90 days
203.
by alansaber 2mo ago
>most popular harness writes a blog about harnesses
204.
by alansaber 2mo ago
You don't need to. Underlying agent operations (how it compacts, utilises memory, etc) is not user navigable.
205.
by alansaber 2mo ago
Certain fundamental technologies are national security. If you proport to be a superpower you can't have your nation be fully reliant on say the Mac OS. Same with agricultural imports, IC development, dot dot dot
206.
by alansaber 2mo ago
>every llm product that isn't a claude endpoint is going to be what you stuff into the claude etc endpoint
207.
by alansaber 2mo ago
Sol is like this too...
208.
by alansaber 2mo ago
If I had to guess, increasingly long model output is either 1. Deliberate choice to make LM output sound more advanced 2. It improves performance of inter-agent tasks
209.
by alansaber 2mo ago
Agent spam is definitely an excercise in diminishing returns
210.
by alansaber 2mo ago
Hence why the UK govt wants to ban VPNs and have our own lil internet firewall

More ›