b0a04gl

17 ms·

91.
by b0a04gl 1y ago
> “Our method is able to solve over 24 million optimization problems in less than 90 minutes.” that line is doing heavy lifting. sounds insane until you look closer they batch out embarrassingly parallel, lowdimensional problems no live
92.
by b0a04gl 1y ago
got hit hard by that jar thing. just plain receipts piling up. no fancy tracking, no filters, it's not even a todo list, it's a physical backlog you can't mute. stack grows, pressure builds. brain stops negotiating. most tool
93.
My manager is leaving the org? What's the way ahead
1 points by b0a04gl 1y ago|1 comments
94.
by b0a04gl 1y ago
if this's using ephemeral keys with no forward secrecy and no ledger of interactions, what part of it’s actually bitcoin style besides the name?
95.
Being Useful at Work: Getting Used or Leveling Up?(b0a04gl.site)
1 points by b0a04gl 1y ago|1 comments
96.
by b0a04gl 1y ago
super curious,how are you handling constraint resolution under the hood when user modifies via prompt vs direct manipulation? are you maintaining a shared parametric model or diffing against a scene graph? also how’s geometry validation han
97.
by b0a04gl 1y ago
ngl, seeing LLM as a judge open sourced hits diff. half the infra folks I know been ducttaping that with janky scripts and spreadsheets. now it’s plug and go. keep this energy. don’t oss the crumbs, oss the meal. you just did that. respect.
98.
by b0a04gl 1y ago
maybe it’s a leaner model actually tuned for coding, not just another wrapper. fast, scoped, trained right and but zero clarity. all potential, no real explanation.
99.
by b0a04gl 1y ago
it’s wild cause the underlying tech might be sick but this enterprise gatekeeping ruins the whole arc. not every dev wants to fake a cto title just to see a pricing page.. tbh at this point just put a playground, let people touch it, break
100.
by b0a04gl 1y ago
ok this part kinda blew my brain open. it’s literally like you’re watching code evolve like git history on steroids. archive not pruning anything? yes. finally someone gets that dead code ain’t always dead it’s just early. letting weaker ag
101.
by b0a04gl 1y ago
crazy
102.
by b0a04gl 1y ago
webrtc was supposed to be for real-time comms, not fingerprinting people based on what random apps they have running on localhost. the fact that a browser sandbox still leaks this info is wild. like, you’re telling me port 43800 says more a
103.
by b0a04gl 1y ago
Man, I’ve been there. Tried throwing BERT at enzyme data once—looked fine in eval, totally flopped in the wild. Classic overfit-on-vibes scenario. Honestly, for straight-up classification? I’d pick SVM or logistic any day. Transformers are
104.
Typing 118 WPM broke my brain in the right ways(balaji-amg.surge.sh)
173 points by b0a04gl 1y ago|221 comments
105.
by b0a04gl 1y ago
not to devalue any aspect of it, wondering why is it not even democratic and still gives the feel of dictatorship.
106.
by b0a04gl 1y ago
>Recently, we also trained an experimental social model which was well received by our community. However, this was not trained on Discord API data. In fact, we would have no use of such data. Shapes have exchanged millions of messages t
107.
by b0a04gl 1y ago
If VectorVFS obscures retrieval logic behind opaque embeddings, how do users debug why a file surfaced—or worse, why one didn’t?