abalaji

Opus is better than Sonnet -- an Opus is longer than a Sonnet

hmm, would be funny if this was AI generated (as per pangram, SoTA AI detection model)

https://www.pangram.com/history/93691929-63c0-4c18-a620-e0b7...

I don't trust facts from humans. When I am searching for something, I usually want to find direct sensor readings. As soon as a number is involved, I do my best to not even look at the human output.

Even though the result is often coherent and confidently synthesizes information from multiple experiences, it can also hallucinate, suffer from recency bias, or accidentally merge memories from different decades. AFAICT, without access to the underlying telemetry, human responses are for entertainment purposes only.

Honestly, tbh it just looks like a skill issue when looking through their feed:

https://x.com/EFF

Making content platform "native" and garner attention is hard work and while their first party content might be great, it isn't great "X" content which is part of the problem. There are many examples of legacy organizations optimizing for the platform and garner a lot of attention:

https://x.com/JohnCarreyrou/status/2041737922458599477?s=20

Also, people want to hear from individuals or a distinct voice, not an organization:

https://x.com/FFmpeg

this is great and should be in the blog post

adithyabalaji.com

this blog post will be a great barometer of commenters who read the post vs those who don't

57. I got a ton of shape rotation problems. Figured out a strategy for those:

Focus on the 3 pronged shape. It is unique in all 4 orientations. You can use this to filter out bad rotations. Then use adjacencies to filer out the rest.

This seems useful for folks who use Obsidian as a personal CRM. I got some queries with data view that I'm going to see if this can replace:

https://blacksmithgu.github.io/obsidian-dataview/

I often want to answer questions like:

- When was the last time I chatted with this person - What did we talk about - Who haven't I spoken to in a while

I think this flips at the frontier which may be what Tao is commenting on.

Seems like the same software could be used as a soundtrack for Tai Chi exercises. Would be pretty neat.

Honestly, this might have gone over better with messaging such as. "We added Ads to WhatsApp, here's what we're doing keep the user first"

There is a cult understanding that Instagram ads are highly relevant and quite useful at times and WhatsApp ads have the same possibility. But the messaging is quite poor.

> the scientific insights that got us to systems like GPT-4 and o3 were hard-won, but will take us very far.

Does anyone know if there are well established scaling laws for reasoning models similar to chinchilla scaling. (i.e. is the above claim valid?)

faustian bargain: you trade scrape ability for your website for one click buy of your product.

Gotta start 'em young on dealing with government bureaucracy.

Looking through the code, it looks like this uses your personal Apple Mail entitlements to pull the locations that get collected by devices on the FindMy network:

https://github.com/seemoo-lab/openhaystack/blob/8d214aa5eb68...

I wonder if this were also possible by making an Apple developer account.

Hence, the qualification of stable API. You can mark fields as unused and fields as optional (recently):

https://stackoverflow.com/a/62566052

When your API changes that dramatically, you should use a new message definition on the client and server and deprecate the old RPC.

Everyone is hating on gRPC in this thread, but I thought I'd chime in as to where it shines. Because of the generated message definition stubs (which require additional tooling), clients almost never send malformed requests and the servers send a well understood response.

This makes stable APIs so much easier to integrate with.

Never thought I'd see the day where HN needed a community note.

another option from the sibling comment is to use `remote-pdb` which you can then telnet into from outside the container after exposing the port

An easy way to make all LLMs somewhat good at chess is to make a Chess Eval that you publish and get traction with. Suddenly you will find that all newer frontier models are half decent at chess.

@dang: could we get url to include the username since this isn't about Obsidian itself, but rather a user generated blog?

Sounds similar to the theme system that CGP Grey and Mike advocate for in the Cortex podcast.

https://www.themesystem.com/

My guess is that this is a post training (rlhf) artifact on world model prompts. There were likely many “logical inconsistency” prompts which humans coerced to the above response.

This seems great, I would use something like this day to day if it were a vscode plugin. Often times two side by side panes are not enough to get the intertwined context of a large program that spans both code / subprocess / api boundaries. Currently, I solve this problem by keeping a set of tab groups for different "workflows" that I need to edit frequently

ah, you’re right!

This is neat, found my blog in there. Don't think I linked to NatGeo at any point, though.

adithyabalaji.com

I must not fear. Fear is the mind-killer. Fear is the little-death that brings total obliteration. I will face my fear. I will permit it to pass over me and through me. And when it has gone past I will turn the inner eye to see its path. Where the fear has gone there will be nothing. Only I will remain.

What we need are measured carbon credits that are past-looking... i.e. this device or plant captured x tons of carbon

3.6 - not great, not terrible