btown

10 ms·

151.
by btown 6mo ago
> ended up enabling groq For those reading fast, this isn't a reference to SpaceX's Grok, this is Groq.com - with its custom inference chip, and offerings like https://groq.com/blog/introducing-llama-3-groq
152.
by btown 6mo ago
I think there's a broader indication of an arms race between noise cancellation systems and things that want to be heard, like advertising. And this just-happening-to-exist bandpass that the DuoBell is depending on could easily become
153.
by btown 6mo ago
Right? This feels like an "arms race" similar to scraping vs. anti-scraping; countermeasures will be developed, likely due to the action of actors entirely disconnected from what you're doing, but to block something else in t
154.
by btown 6mo ago
Are the Gnome features planned to be ported to macOS? Frozen columns and cancelable queries are pretty vital things!
155.
by btown 6mo ago
PWAs were more than an experiment - they were even mentioned in Apple keynotes (IIRC). And sandboxing was every bit as stable as website sandboxing. They were killed because app store operators realized they bypassed an ability to police pa
156.
by btown 6mo ago
The same way coding agents don’t replace the need for an IDE, content generation needs to support arbitrary human-to-agent handoffs, where the human can say “this is the wrong direction, I sketched this change of what I want it to look like
157.
by btown 6mo ago
FWIW this was the status quo (webpage could ping arbitrary ports but not read data, even with CORS protections) - but it is changing. This is partially in response to https://localmess.github.io/ where Meta and Yandex pix
158.
by btown 6mo ago
For what it's worth - and I'm not saying that LinkedIn is doing this for the right reasons - I can imagine a frontend QA team wanting to do this to understand how prominent certain extensions are for users of various parts of thei
159.
by btown 6mo ago
One of the things that LLMs "excel" at, pun very much intended, is this exact pattern - creating filtered aggregates for a finite set of columns, and using this at the end of a CTE! OP's example, for reference, was: SEL
160.
by btown 6mo ago
This is an amazing analysis! Thank you for running this :)
161.
by btown 6mo ago
It’s only a losing strategy if you assume everyone universally adopts the slow strategy, and no research teams spot it in the interim. For things with large splash radius, that’s unrealistic, so defenders have an information advantage. Ma
162.
by btown 6mo ago
It seems the benchmarks here are heavily biased towards single-shot explanatory tasks, not agentic loops where code is generated: https://github.com/drona23/claude-token-efficient/blob/main/... And I thi
163.
by btown 6mo ago
Is there any good documentation about contracts? https://en.cppreference.com/w/cpp/language/contracts.html is incredibly confusing - its first displayed example seems to be an edge case where the assertion it
164.
by btown 6mo ago
If anything, Sora was an experimental question: giving away video generation is expensive, but is the voluntary user labeling and engagement data, which can be fed into RLHF, accretive enough to model training that it's a meaningful tr
165.
by btown 6mo ago
Gosh, I wish this had existed a year ago; I spent an absurd amount of time creating a system for print brochure typesetting in HTML, that would iteratively try to find viable break points (keeping in mind that bullets etc. could exist at an
166.
by btown 6mo ago
At least, not that we remember.
167.
by btown 7mo ago
I wholeheartedly disagree with this. For any iteration, Claude should be reading your codebase, reading hundreds of thousands of tokens of (anonymized) production data, asking itself questions about backwards compatibility that goes beyond
168.
by btown 7mo ago
I agree with everything you said - but it's also the case that a set of parameters were created that, instead of requiring multi-person validation of target validity and provenance, prioritized speed to provide decision makers with opt
169.
by btown 7mo ago
I think that three things can simultaneously be true: (1) missile defense systems based on deep data fusion with cutting-edge espionage systems for launch detection are becoming increasingly useful and necessary (2) we should be thoughtful
170.
by btown 7mo ago
Your point is well taken, though it's worth pointing out that literally yesterday Palantir was co-awarded a contract for building orbital weapons systems [0]. The broader point is Palantir's specific confluence of: - access to gra
171.
by btown 7mo ago
I do think that what we think of as RAG will change! When any given document can fit into context, and when we can generate highly mission-specific summarization and retrieval engines (for which large amounts of production data can be held
172.
by btown 7mo ago
> That’s why selling SaaS or AI to this kind of company isn’t for me - I’d rather focus my energy on building a company from my own principles, and hire people who share them from the beginning. > When I told my manager I was leaving,
173.
by btown 7mo ago
One very interesting instance of the "military data moat" is Ukraine's annotated database of drone footage, perhaps the first of its scale from live engagements [0]: > They can now draw on an enormous pool of real warfare
174.
by btown 7mo ago
The economic effect of latency is measured not by the incremental productivity itself, but by the combined economic downforce of thousands of resulting HN and Reddit comments :)
175.
by btown 7mo ago
Throughput is indeed king for the standard-tier mindshare-capture play. But there are many who would pay multiple times the current cost for agentic systems for engineers and executives, if it meant a meaningful reduction in latency. The ec
176.
by btown 7mo ago
It also seems worth considering that the article's view that "spending a lot of time searching for the carrier is a good way to get killed by defending fighters" is a distinctly pre-drone-ubiquity assumption. Can a carrier gr
177.
by btown 7mo ago
Neat! In case you took me too literally: railroad diagrams are fun, but far from the only way to give spec level clarity, so don’t feel you need to overindex on my silly comment! I am curious why it’s parsed right to left. Is this so that y
178.
by btown 7mo ago
This is really interesting. At first glance, I was tempted to say "why not just use sqlite with JSON fields as the transfer format?" But everything about that would be heavier-weight in every possible way - and if I'm reading
179.
by btown 7mo ago
There was another comment I recall from today, discussing how OpenAI is not-so-subtly adopting a social-network-esque model, in how it's fine-tuned its chat system to always suggest another question that the user might want to ask. And
180.
by btown 7mo ago
I recall that this became a big problem for the Homebrew project in terms of load on the repo, to the extent that Github asked them not to recommend/default-enable shallow clones for their users: https://github.com/Home

More ›