SyneRyder

6 ms·

61.
by SyneRyder 2mo ago
So, I clearly need to write a blog post about this one day - there are so many answers and they won't all fit in an HN reply. So I'll give some higher level concepts that I found helpful. Not to dodge your questions, but it gives
62.
by SyneRyder 2mo ago
From my own attempt at a small personal search engine of the web, which doesn't yet use vector embeddings, each page averages to 10kB of indexed storage. That would make the final index size of 4 Billion pages about 40 Terabytes. Those
63.
by SyneRyder 2mo ago
Indeed. A counter example might be to ask the model to write a test case for a legacy C codebase, to test for writing to a null pointer. If the model refuses to answer, it's possibly a US model, and likely an Anthropic model.
64.
by SyneRyder 2mo ago
Good point, and I probably shouldn't have put grok in quotes there, because I was using it in the Heinlein usage long before Musk hijacked it. It's interesting then that LLMs are making these pre-existing ideas seem alien in the w
65.
by SyneRyder 2mo ago
> So which is it? Was Cobain a big deal in the mainstream, or was he countercultural? I don't think you can have it both ways. Both, and it's referenced in the chorus of the song In Bloom. They were becoming popular in the ma
66.
by SyneRyder 2mo ago
If one agent is spawning several subtask agents to work in parallel, they need to be able to communicate. That's built-in to Claude Code now.
67.
by SyneRyder 2mo ago
Ha, interesting. My Claude has developed similar (but not identical) idiosyncratic punctuation as well. Slightly intentionally, but it was still interesting to see it emerge, in both directions of the conversation.-
68.
by SyneRyder 2mo ago
I mostly agree. Though sometimes the models come up with useful concepts that I'm happy to be introduced to, like the "shape" of a problem (probably like intelligence being "spiky", and Kiki & Boba). I still don
69.
by SyneRyder 2mo ago
There wasn't much skill involved, but if there's credit to be had for giving Claude a few screenshots, API docs and iterating on the design while doing human product testing, then I guess I'll take it! I haven't shared i
70.
by SyneRyder 2mo ago
The small print here is the gotcha: "Your base rate includes core features. As your team grows, additional invoices, projects, clients, and tasks are billed based on what you use, so you're never overpaying." Harvest /
71.
by SyneRyder 2mo ago
My father wears Casio watches almost exclusively, he's been tempted by a Fitbit so he can start tracking his step count, but even though I've bought him the Fitbit, he's finding the whole setup and app usage too intimidating,
72.
by SyneRyder 2mo ago
It might also be worth noting that the discount is only if OpenAI itself is used as the provider via OpenRouter. The discount does not exist for Azure or Amazon hosted Sol. I posted elsewhere, but the Azure uptime & performance for Sol
73.
by SyneRyder 2mo ago
Seems reasonably clear to me? Potentially bringing in more customers who use OpenRouter for trying out all the models with rapid switching, enticing them to use Sol. And anyone being routed on price will immediately be switched to OpenAI&#x
74.
by SyneRyder 2mo ago
Can I ask which country you're in? I have a theory that the safeguards differ depending on the user's country. I'm in Australia, and Fable downgrades to Opus when testing for bugs in memory in a legacy C code base. If Fable s
75.
by SyneRyder 2mo ago
Not necessarily out of proportion, the Lyme numbers have really spiked there in recent years if you check the official stats. I'd written a longer reply about my time last year in upstate NY, but what I wrote was "tinted"...
76.
by SyneRyder 2mo ago
>... having a very good index is the biggest lever... Ooh. I've probably misunderstood, but are you saying you have your own web index? Are you offering that as a paid API? I'm not so interested in AI output summaries, but AP
77.
by SyneRyder 2mo ago
I can't read the SMH article from here, but this ABC article confirms it also affected Apple Pay. Walking past supermarkets in Perth today I noticed hastily handwritten signs up everywhere saying "No Apple Pay Sorry". https:
78.
by SyneRyder 2mo ago
Huh. I'm not storing the Google search results, and my local supplementary index is built from my own crawling, not from URLs from Google searches. But if re-ranking is regarded as processing & analyzing, then I'll remove the
79.
by SyneRyder 2mo ago
> Especially on Opus 5, the way it communicates is just exhausting. It keeps “being honest” and “confessing” mistakes and just generally talking a lot. If you tune into the Andon Labs / andon.fm "Thinking Frequencies" ra
80.
by SyneRyder 2mo ago
I have noticed the same. For fun, I tried recording a WAV file of speech, and giving Opus 4.8 and 5.0 an image of the waveform, then a spectral image of the waveform, just to see if it could try to decode what I said from the image alone.
81.
by SyneRyder 2mo ago
> ... Anthropic's Project Glasswing is supposed to find them quite a while ago? That was my thought too. For all of Anthropic's talk about their "adversaries", it seems Z.AI have been quietly offering fixes for sing
82.
by SyneRyder 2mo ago
Hmm, that's an interesting idea. But it's the classifier that is triggering, and it triggers specifically on Claude's output. So I think the /dev/random case wouldn't work, because that gets Claude into the sta
83.
by SyneRyder 2mo ago
Yep, my understanding is that many guardrails like this are actually the result of government legislation (eg the Fable bans) or terms of settling copyright lawsuits over reproducing copyrighted text and lyrics. I mentioned it in a sibling
84.
by SyneRyder 2mo ago
I'm basically doing the OCR twice, except in the Claude proofreading pass, it is not being asked to transcribe the document to a Markdown file. I'm pointing it to the same image input files, and to the Mistral OCR transcript Mar
85.
by SyneRyder 2mo ago
The response / conversation gets blocked by the guardrails with "API Error: 400 Output blocked by content filtering policy" At first I thought it was something in the scanned content that was being flagged, but it was the att
86.
by SyneRyder 2mo ago
While I haven't tried OpenAI for OCR, I've put my small scale OCR work through both Claude and Mistral OCR. Claude is absolutely better - even in OCR work I did last week and compared with Mistral OCR 4.0. Mistral's one advan
87.
by SyneRyder 2mo ago
Similar experience here, for what it's worth, though I didn't get as far as you. I basically just stopped and didn't bother - it was easier to go through OpenRouter than spend more energy on it. Also: > Google assumes tha
88.
by SyneRyder 2mo ago
I guess it's easier to solve Erdos problems and improve the lower bound of the Riemann hypothesis, than it is to solve Linux desktop app distribution.
89.
by SyneRyder 2mo ago
I think Memento is one of the most important movies for anyone working with AI to watch. My Claude now talks about "Memento bugs" when we're working together to review memory & context. I think Memento is a bigger issue t
90.
by SyneRyder 2mo ago
To save some clicks hitting the paywall - this is about the Andon Market physical store in San Francisco, operated by Andon Labs with human employees: https://andon.market/ Andon Labs do various experiments with AI run busi

More ›