hedgehog

4 ms·

61.
by hedgehog 2mo ago
You've made a really sharp observation, and the reason it lands is worth naming:
62.
by hedgehog 2mo ago
On a brief background read it looks like another issue is that blobs are not necessarily stored in contiguous regions of the file. Maybe there's some clever way to combine a VFS shim and virtual tables to store page aligned blobs in a
63.
by hedgehog 2mo ago
I didn't understand the core issue, is it that due to non-alignment of the on-disk data you have to do a copy at load time, but if you could guarantee alignment of the actual blob content then you could use it directly?
64.
by hedgehog 2mo ago
It's not hard to make a portable UI framework if you accept similar limitations to the terminal (limited OS integration, accessibility support, etc). You can do anything you want and ship it through RFB.
65.
by hedgehog 2mo ago
Prime Agent looks really interesting. Both the "recursive language model" bit and routing everything through IPython. https://github.com/PrimeIntellect-ai/prime-agent
66.
by hedgehog 2mo ago
Multi-turn, expose a tool that extracts crops. Works great.
67.
by hedgehog 2mo ago
I don't know how the contemporaneous IRIS would compare for the target workloads but not necessarily faster than a stock NeXT, let alone one with whatever upgrades the customer wanted to stick in there. The 68000 and by that point the
68.
by hedgehog 2mo ago
Factory overclock, and in those days Motorola could probably have given them a batch of mid-development DSP chips (sort of proto 56002) that could clock much higher and still sit on the unmodified motherboard, max out the scratchpad memory,
69.
by hedgehog 2mo ago
I was pretty disappointed that didn't work out, but it also is a direct precursor to Mojo which actually makes me more optimistic about Mojo's design choices.
70.
by hedgehog 2mo ago
I think a bigger issue than open source timing is the community governance necessary to shape a tool that serves the needs of a lot of stakeholders. Swift has lived under the shadow of Apple's (often secret) product priorities. The ext
71.
by hedgehog 2mo ago
Methane has a stronger greenhouse effect than CO2 (until it decays into CO2) so it's probably better to burn it and do no work than just vent it, and better still to leave it in the ground.
72.
by hedgehog 2mo ago
Log all the calls and run on a periodic cadence (cron or ever N turns) a larger model (like Opus) to read samples of the traces and edit the template to fix observed problems. There are some signs that help find interesting things to look a
73.
by hedgehog 2mo ago
It's a bit of a different situation though, in many cases the unit of adoption is a team or organization that is free to switch platforms without consideration of what anyone else is doing. At this point I'm going to move my perso
74.
by hedgehog 2mo ago
To be fair a lot of models have quirks, I've never found a model swap that was transparent.
75.
by hedgehog 2mo ago
Look for improved templates by "froggeric" on Hugging Face. I use Qwen 3.6 a fair amount using his template and it fixes some issues I saw with the upstream versions.
76.
by hedgehog 2mo ago
This is all a pretty active area of research, both adaptive drafting, different draft approaches (DFlash etc), I expect to see really big improvements land in the popular runtimes this year.
77.
by hedgehog 2mo ago
Yes, I was fixing issues piecemeal until I found the froggeric template, I've had to fix I think one issues with that one but it's better.
78.
by hedgehog 2mo ago
Same, I have one workload where on 3.6 drafting 6 tokens is the fastest setting.
79.
by hedgehog 2mo ago
I think the idea is you do visual editing on your page as rendered, and then Pixy generates a log of what you did annotated in a way a coding agent can use to generate code changes. Seems convenient.
80.
by hedgehog 2mo ago
I have some self-improving harness tooling, I run it within Claude Code because of the cost advantage but that does mean restarting Claude Code periodically so it can pick up all changes. It sounds like the DeepSeek system won't need t
81.
by hedgehog 2mo ago
The fun part is unless your app is pretty high traffic you might see the bug only once, or twice, and never have a reasonable way to even get close to a fix or know what your real exposure is.
82.
by hedgehog 2mo ago
See sibling answer but essentially the effectiveness of cache is not diminished by having a separate one per model (relative to the win of doing more turns and generation with a cheaper model).
83.
by hedgehog 2mo ago
To elaborate, because I don't think some of the people reading this understand the reason, typically a lot or most of the cost in "agentic" API usage is cached read + generation. Cached read costs scale with turn count, which
84.
by hedgehog 2mo ago
As pjmlp points out we don't know what would have happened if Intel had spent another few billion dollars re-implementing IA64 in a lower cost / lower power form. To me one of the big lessons of the last few decades is that pragma
85.
by hedgehog 2mo ago
You can set it in the config file so it's the default every time.
86.
by hedgehog 2mo ago
It's unclear to me what would have likely happened had AMD stayed on IA32 and Netburst been uncontested. Itanium was far too expensive for consumer machines and it seems very possible that Intel would have fumbled IA64 in a way similar
87.
by hedgehog 2mo ago
They Toyota hybrids are still gasoline, what's the objection to the eCVT? So long as we're still using combustion engines it seems like the best design anyone's come up with.
88.
by hedgehog 2mo ago
https://archive.is/kaef0
89.
by hedgehog 2mo ago
They are useful for packaging all the related bits together. Some instructions on the line to add to the main prompt, detailed policy doc that spells out details, troubleshooting docs, the code itself, etc.
90.
by hedgehog 2mo ago
I like it based on initial testing today. I already use my own local UI for managing stacked PRs so I can see the dependencies as a tree view and the review + CI status for each, it would be good to also have those in the GH web UI. Maybe I

More ›