jawon

I got Opus to knock out an MCP server that implements subagents running in pi and tell Opus to send work to DeepSeek. Or I tell it to ask GPT-5.5 for critiques. It's manual but saves a lot of tokens.

Also, from a comment in another thread, from roon, who works at OpenAI:

> codex-5.2 is really amazing but using it from my personal and not work account over the weekend taught me some user empathy lol it’s a bit slow

[0] https://nitter.net/tszzl/status/2016338961040548123

Not magical thinking, not conspiratorial, just hypothetical.

Just because you can't afford to 10x all your customers' inference doesn't mean you can't afford to 10x your inhouse inference.

And 2.5x is from Anthropic's latest offering. But it costs you 6x normal API pricing.

I was thinking about inhouse model inference speeds at frontier labs like Anthropic and OpenAI after reading the "Claude built a C compiler" article.

Having higher inference speed would be an advantage, especially if you're trying to eat all the software and services.

Anthropic offering 2.5x makes me assume they have 5x or 10x themselves.

In the predicted nightmare future where everything happens via agents negotiating with agents, the side with the most compute, and the fastest compute, is going to steamroll everyone.

This is “why are we going to space when we haven’t cured cancer” reasoning.

I'm building my take on a low-touch task completion assistant designed to counter distraction and hyper-habituation.

It's starting off as a MacOS app because that's the machine I have. I didn't know Swift or SwiftUI when I started. I now know them somewhat, but the entire app has been vibe-coded. This has made it slow going. Very "1 step forward 2 steps back" until I switched from Claude Code to Codex and GPT-5.

I'm hoping to start an initial beta within the family in the next week or two, and then a wider round in January.

Piracy might not hurt sales, but 1000 publishers putting out their own copies of your book/game/song/poster/miniature once it hits the market will.

That's why I can accept copyright even thought it's not perfect.

Are these numbers full time employees only or total FTEs? Because it mentions Walmart: "Walmart’s full-time employees number remained relatively constant for the last 10 years".

Would revenue / person-hour show a different trend? Because there are a lot of part-time and contract workers out there.

Guanfacine.

I found it made me think and act sleep deprived even though I didn't feel it and also increased anxiety.

Guanfacine is also an alternative, and it's method of action also makes it anxiety reducing.

You might be how surprised how low a dose you need for an effect. 5-10ug of Ritalin noticeably reduces the "noise floor" for me.

How do you take 5-10ug? Dissolve 10mg in a litre of something. Get a 1ml dosing syringe. It has 0.1ml markings.

You could start there and increase it until you find what works. Also, if you take very little you can have a break on weekends and not suffer too much while remaining sensitive to lower dosages.

This is a standard book copyright notice:

All rights reserved. No part of this publication may be reproduced, distributed, or transmitted in any form or by any means, including photocopying, recording, or other electronic or mechanical methods, without the prior written permission of the publisher, except as permitted by U.S. copyright law.

“Reproduced” and “electronic” are the relevant terms here.

I remember when gpt-3 came out and you could get it to spit out chunks of Harry Potter and I wondered why no-one was being sued.

The models are built on copyright infringement. Authors and publishers of any kind should be able to opt out of being included in training data and ideally opt-in should be the default.

And I hope one day someone trains a model without the use of works of fiction and we find a qualitative difference in their performance. Does a coding model really need to encode the customs, mores and concerns of Victorian era fictional characters to write a python function?

What size gemma are you using? Is the refactoring running independently or managed by you?

Gave it a try. After a few minutes I felt more like I was recognising the samples than I was recognising the notes. Not sure what you can do about that short of physically modeling an instrument.

This definition of AGEs is interesting (linked from the article, same site):

"Advanced glycation end products (AGEs) form when proteins and fats (lipids) in the body react with sugar (glucose) and become glycated and oxidized."

So, maybe it's not so much the Maillard reaction but that sweetened brioche bun used in your fancy hamburger.

I like msty.app. Parallel prompting across multiple commercial and local models plus branching dialogs. Doesn’t do artifacts, etc, though.

Interesting. I have the exact opposite issue. Hypervigilance and all that.

People out there trying build some semblance of AI out of an LLM using larger and larger networks of “agents” that generate, classify, revise and verify data using the same LLM they're building larger and larger networks of agents upon to try and build some semblance of AI.

The end game is a brain-sized network where each neuron is an agent sending a 1M token prompt to a 10T parameter model to update their "weights".

I've never seen this before. I love the design. It's like a drum machine for spreadsheets. And it's from 1983, the same year the Tandy Model 100 was released. A good year for gadgets.

"The team used mice that develop a form of Alzheimer's. They exposed these mice to bursts of sound and light that occurred 40 times a second.

The stimulation induced brain waves in the animals that occurred at the same, slow frequency.

Tests showed that the waves increased the flow of clean cerebrospinal fluid into the brain and the flow of dirty fluid out of the brain. They also showed that the fluid was carrying amyloid, the substance that builds up in the brains of Alzheimer's patients."

Does this mean we might be able to use something like TMS instead of sleeping?

And could a failure or reduced functioning in this system explain those people "allergic" to electromagnetic radiation?

I look at all the advice on prompts and I don't feel like the "Frame Problem" has been solved. It feels like it has shifted into the "Frame Invocation Problem".

And it is this very problem which led me to ask my question about different architectures.

Thanks for that. It looks like the kind of thing I'm looking for. I'll give it a read.

Can anyone point me to models that look like they might actually be useful in moving towards AGI? I feel like I have a basic understanding of the transformer architecture, and multiplying X tokens in a sliding window across a set of static matrices to produce 1 new token does not look like a path to AGI.

Yes, the complex feature extraction is impressive. But are there any models that, I don't know, are more dynamic? Have a working memory? Have a less limited execution path?

But it's just spent 30-45 minutes at boiling point. If you keep the cooker closed and only use clean utensils to remove rice I'd say the risk is pretty low.

I assume it's not 100 billion floating point numbers sitting on an SSD. Not even if they are cleverly filtered to select the best 4096 of them.

I really wish someone who claims an LLM is conscious would explain how the consciousness is being generated while walking through the code that generates the text to support their claim.

Like, is it some kind of giant pool of static experience that is being "accessed" via the context window? What's the "theory of mind"?

They know the weights aren't being updated between sessions, or during sessions, right?

I do something similar to Flowtime but without the bookkeeping: Orodomops.

Set a stopwatch going but out of sight. When you run out of steam check it. If it’s been less than 25 minutes keep going. If greater, have a break scaled to the time you worked.

I started doing this because I couldn’t get on board with breaking up the work I enjoyed doing, but I recognized the utility of having breaks when that wasn’t the case.

What impact are you hoping it will have? If you're worried about losing your flame try micro-dosing it. I've found < 10 micrograms of ritalin can have a useful impact - gets rid of some of the background noise but without the negative effects of too narrow a focus.

What does "vitamins through fenugreek" mean? And are you just eating regular goji berries or using an extract?

I did something similar on Metafilter - a Greasemonkey script to hide comments with a low number of favourites - and they were outraged.