jbellis

21 ms·

121.
by jbellis 1y ago
Looks like this is targeted more at "better autocomplete" than "powers your next agent." Makes sense given Apple's interest in the on-device experience.
122.
LLMs and Artists(blog.brokk.ai)
2 points by jbellis 1y ago|0 comments
123.
by jbellis 1y ago
The search defaults are pretty bad, but I'm happy with "fuzzy" mode.
124.
by jbellis 1y ago
I guess if you love writing DSLs this is an unfortunate development, but for me it's more of a glass half full: I can have the AI spit out boilerplate I need to solve a problem instead of spending a week building a one-off DSL compiler
125.
by jbellis 1y ago
Only if you can't read. DSv3 is 4x cheaper than the cheapest option in that link during US business hours.
126.
by jbellis 1y ago
Love to see it, this takes Flash Lite from "don't bother" territory for writing code to potentially useful. (Besides being inexpensive, Flash Lite is fast -- almost always sub-second, to as low as 200ms. Median around 400ms
127.
by jbellis 1y ago
I think this is correct, and I also think it holds for reviewing human-authored code: it's hard to do the job well without first having your own idea in your head of what the correct solution looks like [even if that idea is itself fla
128.
by jbellis 1y ago
so basically litellm proxy?
129.
by jbellis 1y ago
and all of them are so much more expensive than OG deepseek as to completely remove themselves from consideration you should probably use grok 3 mini if you want "cheapest model that is reasonably good at code"
130.
by jbellis 1y ago
you're right, I should clarify that I'm talking about no thinking mode, otherwise flash goes from "a bit more expensive than dsv3" to "10x more expensive"
131.
by jbellis 1y ago
no, but if you're not latency sensitive you should probably be using DeepSeek v3 (cheaper than flash, significantly smarter)
132.
by jbellis 1y ago
This is a useful analysis, but only as a first cut and sometimes not even that -- Grok 3 mini and DeepSeek V3 are by far the least expensive coding models that are worth trying for scenarios where you do and don't care about the vendor
133.
by jbellis 1y ago
Did it get upgraded in-place again or do you need to opt in to the new model?
134.
by jbellis 1y ago
Yes! This is an underappreciated point that both sides in "AI makes coding better" vs "AI just writes slop" have mostly missed. Co-authoring code with AI is a qualitatively different activity than writing code by hand,
135.
by jbellis 1y ago
cool, TIL! > Examples of Vineflower's output, compared to other decompilers, can be found on the wiki. [wiki is empty] :-/
136.
by jbellis 1y ago
I don't think it's available in a standalone repo but it IS available as a standalone library, IntelliJ's FernFlower decompiler is the gold standard https://github.com/JetBrains/intellij-community/bl
137.
by jbellis 1y ago
impressive, but that's 1/5 to 1/10 of the throughput that you'd get with a hosted provider, with 1/4 to 1/8 the supported context
138.
by jbellis 1y ago
gc is negligible impact on llm agents the others ("zero dependencies") are not actually related to efficiency
139.
by jbellis 1y ago
I'm surprised that performance wasn't on the list. The main reason I started Brokk on the JVM was for access to https://github.com/joernio/joern , but high performance code that can easily go multicore is a ve
140.
by jbellis 1y ago
sort of, there was the original bigann and then they followed up with a couple more specialized contests the following year, i think it's over now ~300M modern-sized vectors is pretty close to jvector's limit in a single index (th
141.
by jbellis 1y ago
"for android and ios" imo the article's rather negative tone is not justified here, when I only have a small screen I think I would welcome an AI summary in any case I'm willing to see what it looks like before panning i
142.
by jbellis 1y ago
modern dimensions, yes mixed workloads, also yes, especially in an "online" environment rather than the "batch mode" that ann-benchmarks does today but most importantly, multicore -- ann-benchmarks is limited to a single
143.
by jbellis 1y ago
love to see it, there's a huge opportunity in marrying this kind of traditional static analysis with AI tooling like, literally years' worth of research that has mostly been ignored by industry because it's too complex /
144.
by jbellis 1y ago
As the author of a vector search engine I was low key excited for this (there is no good benchmark for vector search out there that resembles real world use even a little, all the vendors have their own internal stuff) but I think using the
145.
by jbellis 1y ago
ahh, that makes sense, thanks I do think that calling this a CAPTCHA when it's not actually intended to distinguish humans from computers is a bit misleading, but I can see why you would do that
146.
by jbellis 1y ago
interesting experiment! it does seem like the consensus is that o3 and (at a much lower cost) r1 are better writers than claude, but obviously anthropic's agent framework doesn't support those
147.
by jbellis 1y ago
But Human+Ai is far more productive than Human alone, and more fun, too. I think antirez would agree, or he wouldn't bother using Gemini. I built Brokk to maximize the ability of humans to effectively supervise their AI minions. Not a
148.
by jbellis 1y ago
Adding support for tool calling is expected, and welcome. The old R1 felt smart enough to drive Brokk's Architect and Search agents, but they won't work without tool calling; I'll try out the new one.
149.
by jbellis 1y ago
I hope we figure out how to scale up diffusion models. Gemini is a cool tech demo, and Inception is almost smart enough to be interesting. There's a lot of tasks that you just need a certain intelligence floor for, and after that getti
150.
by jbellis 1y ago
I don't see anything here to indicate it's "actually" Sonnet under the hood Possibly it was intentionally trained on some of Sonnet's outputs, but given that this only happens in thinking mode and Sonnet 3.5 did not

More ›