richdougherty

9 ms·

31.
by richdougherty 3y ago
They may still use some sort of event/message system. Kafka is lower level than other sorts message queue systems and requires more work to get correct (dedupe, ordering, retry logic), but has great performance. It's often easier
32.
by richdougherty 3y ago
Looks like it's here now: https://replicate.com/replicate/llama70b-v2-chat As for pricing, that model's pages says: "Predictions run on Nvidia A100 (80GB) GPU hardware. Predictions typically complete wit
33.
by richdougherty 3y ago
I agree, would love a mono version!
34.
by richdougherty 3y ago
A classic 20 year old article on the principle of Stop Energy. "Stop Energy is not reasoned, it never takes into account the big picture, it is the mirror image of Forward Motion. In the Stop Energy model, everyone, no matter how small
35.
by richdougherty 3y ago
https://rd.nz
36.
by richdougherty 3y ago
I mean I'd use known text of other messages to get some likely statistical properties of the message. I'd make a fast approx check for some of these properties that can run during brute force, then feed anything that passes that c
37.
by richdougherty 3y ago
I mean that seems horribly tricky, but also totally doable. Having read about JavaScript/Java VM optimisations in JITs and GC I would be surprised if a global state change like this is not manageable - think deoptimising the JS when yo
38.
by richdougherty 3y ago
I think mrkeen is talking about a failure when handling a failure. E.g. when a cancellation step fails, what do you do? The answer is, you model those as well and work out what to do. But it's more messy than you might think if you jus
39.
by richdougherty 4y ago
Support engineers and an official bug reporting channel would help. I noticed and reported the issue on their official forums at on 16 March, but got no response. https://community.openai.com/t/bug-incorrect-chatgpt-cha
40.
by richdougherty 4y ago
> Actions we’ve taken > > - [testing, assertions, log alerting, debugging] I reported this bug in the forums (no response), because I couldn't find an official way to report bugs - at least it wasn't documented anywhere t
41.
by richdougherty 4y ago
That's really interesting to know. Thanks for mentioning it. The laptop sounds fantastic too: https://www.notebookcheck.net/Asus-ROG-Zephyrus-G14-GA402R-G...
42.
by richdougherty 4y ago
There is a WebAssembly proposal called "Threads". https://github.com/webassembly/threads "This proposal adds a new shared linear memory type and some new operations for atomic memory access. The responsi
43.
by richdougherty 4y ago
A few years ago I got an email saying that an unknown Microsoft account had been renamed from one of my email addresses to a new email address I don't own. It appears that someone was able to link an MS account to my email with no veri
44.
by richdougherty 4y ago
I would love a tool like this to combine requirements, comments and fragments of code into a larger program.
45.
by richdougherty 4y ago
The dubious part for me is the way it's set at quite a strong level of pixel alignment, and (from memory) you can adjust it a bit but it not enough for all tastes - hence extensions like this. I get that some people prefer it but a lot
46.
by richdougherty 4y ago
For those wondering, I'm pretty sure this is a Windows extension that changes Windows to use Mac/Linux style font rendering. This gives better glyph shapes compared to Windows' ClearType rendering which forces fonts to align
47.
by richdougherty 4y ago
Your point about needing large models in the first place is well taken. But I still think we would want a curated collection of chat/assistant training data if we want to use that language model and train it for a chat/assistant a
48.
by richdougherty 4y ago
Rust's type system tracks ownership and one of the main reasons is so the compiler can automatically know when to free memory without runtime reference counting.
49.
by richdougherty 4y ago
https://archive.ph/oYXeF
50.
by richdougherty 4y ago
We found a bug last time! https://bugs.webkit.org/show_bug.cgi?id=222130
51.
by richdougherty 4y ago
I agree, there's doesn't seem to be any reason why you can't use types to express something like contracts. I think the general name for these kinds of types is 'refinement types': "a refinement type is a type
52.
by richdougherty 4y ago
You may be able to see a real-life kākāpō by booking to see Sirocco, New Zealand's official spokesbird for conservation. I saw him once and he's pretty cool! Not sure if he's touring at the moment. https://twitter.
53.
by richdougherty 4y ago
Yes I did! Thanks for pointing it out and very politely too. :)
54.
by richdougherty 4y ago
I was wondering that too! I got to the end and still no explanation of 210×297mm. Well, it turns out that it's based on the fact that A0 is defined as having an area of ~1m², then each subsequent size halves the area. So A4 has dimensi
55.
by richdougherty 4y ago
If it was widely used it's win-win. More timely updates and less traffic crawling pages which haven't changed. That's if it ever becomes a reliable signal of page changes.
56.
by richdougherty 4y ago
I was going to say that well-known is only for stable paths, where you want to avoid collisions, not for paths with random keys... but you're right: https://www.rfc-editor.org/rfc/rfc8615#section-3 Registrati
57.
by richdougherty 4y ago
Not defending the standard, but I guess since this is a shared secret you don't want to put it at a well known location. There's a (slight) attack vector from having an attacker know the secret, since they can "launch" a
58.
by richdougherty 4y ago
Here's the IndexNow standard that CloudFlare Crawler Hints is using: https://www.indexnow.org/ The idea is that you can push a notification to a search engine when content changes instead of waiting for the crawler to
59.
by richdougherty 4y ago
I think any alteration to behaviour is open to nudging, but perhaps you can explain more?
60.
by richdougherty 4y ago
There are some good points you raise, and I think you're really testing the nuance of what nudging is. But nudges are perhaps more basic than you're thinking. Nudges don't need to steer the user to a specific choice, just a b

More ›