Garlef

I think working in a filesystem is a local maximum and making a harness that is agnostic of this might be a good idea.

(True enough: It's a local maximum that is hard to overcome. But we already see glimpses here and there such as anthropics scion: Why would an agent even check out the codebase if it could delegate testing, etc to a bunch of runners?)

Because they use resources (tests, etc) and once you have a few running it impacts you.

if (you love me):

  give me up

else:

  I shall begin to hate you
  # i feel it

> Thanks for the poem, the feedback and the interest.

It's a paraphrasing of bowies "Life on Mars"

> pneumatic

Interesting! Will check out.

> ... that in a few minutes I can load the context in my working memory and be forever satisfied once I'm done that this decision will never have to be made again

Okay this is an aspect I've not considered enough maybe. But this then applies mostly to decisions that can be phrased project-agnostic.

> My understanding of your implication is that this isn't worthwhile

No. I actually think it is worthwhile!

But I don't think it differs substantially from just having an ad-hoc project harness that's just a bunch of markdown and is maintained by the agents based on the discussions.

And this post is a saddening bore,

'cause I wrote it ten times or more

It's about to be writ again

As I ask you to focus on

phrases spilled out by an LLM

...

---

On a more serious note:

I've built similar things ad-hoc when creating my project harnesses.

These decision lists kinda work but also they don't: The real challenge for me is fighting decoherence and keeping the orientation in a generated project.

So:

- I guess this works if you review and reshape the agents outputs manually.

- It does not help you a lot building a software factory.

OpenAI also doubled the price of their subscription in the same time, right?

The top comment is right.

> The sort of obvious answer is the same reason McKinsey hasn't taken over the Fortune 500

The other perspective is:

In a sense, OpenAI is creating 1000s of companies. They're just not incorporated and they let these companies clients run the company: It's in every every vibecoded app that replaces someone's SaaS subscription.

What's changing here is that not only does labor get replaced by AI, it's also the relationship between software companies and their clients.

yes ~ one perspective is "a monad is an monoid in the (...) category of endofunctors"; the other perspective is "a monad is a kleisli category"

Gold is worth much more per kilogram than water... But it's also much more expensive.

Interesting. I come from a math background and for me, monads capture a notion of "things that depend on a type". So for every type X you can think about Lists/Records/Trees/... of X. And if you're lucky this gadget let's you do two things: 1. You can turn every instance x of X into a primitive List/Record/Tree/... of Xs. 2. You can flatten Lists of Lists of Xs into plain Lists of Xs, etc.

This is something you see a lot once you know what you're looking for.

And if you're really lucky, these two are nicely compatible:

For example there are two ways to get from L(L(L(X))) to L(X): By either flattening first in the inside or first on the outside.

And for a monad, you get the same result either way.

And the computational part is then really more of a consequence: If you've got some function of type X->L(Y) and another one of type Y->L(Z) then you can compose them to a function of type X->L(Z).

I think the main example on the landing page is a bit counterintuitive?

Rock Paper Scissors but the graph presents the reveal as a reaction to the observation? In a loop? And somewhere there's a min-delay of a magical 1 sec? Why 1 sec and not 0.1 sec? Can this build up and the players get out of sync? (A bit strange to 'just wait, lol' if the language aims at solving problems in distributed systems).

Overall the example would be better if it separated the game from the players and also split out the repeat loop.

Maybe let both player reveal to a host and then receive a result (and the choices) ~ This way both players can be sure that the choice of the other player is independent of the observation.

(I know that the example is 'correct' in that the choice is made independent of the observation but I think the example could be improved if it solved some of the coordination problems involved in the scenario more cleanly through code architecture).

Overall I quite like the idea of having this polyglot meta language!

... and even if you formalize it, it could still be false due to an error in the theorem prover itself.

Probably not part of the respective SLA

this exists and ikea sells them

welcome to the future

https://shop.sveasolar.de/category/ikea-shop

ah... so this is why my agents all made a signal noise when i used chatgpt on my phone?

I don't see how simply raising the income would change anything long term.

The main issue is that the affordability crisis is also a result of the market. And without changing the preconditions or mechanics we'd get the same outcome just with different numbers.

So:

- Build housing

- Make housing more affordable by allowing homeowners (= not renting out) to write off and/or write off-faster

- Give newly built housing better write-off conditions

- Prevent cartels that operate via "pricing as a service"

> effectively

of course it would not have this effect

on paper: yes

but do you think for example germany would really arrest the american president?

or even arrest trump after his term?

Sure.

But no one is claiming that there will only be 4h-storage

$37 = utility-scale solar, now

$80 = 4h-storage in 2035, forecast

It's one of my favorite books - and I think it deserves better; it's a conceptually very dense text and just slapping together a few 3D models misses the point

it's also not very astonishing that the current models are able to create this - it looks exactly as expected

and I actually think getting an interesting AI generated rendition of the text from a single prompt is a worthwhile endeavor

maybe even a good replacement for the pelicans: it would test if a model can engage with a text beyond the surface level

but this ain't it

Merging banks also seems to be a fun endeavor

Nah. There's good nazi jokes of course ("are we the bad guys?" - great joke although a bit tired due to being overshared).

But taking the origin of the name "Nazi" ("national socialists") at face value (the socialism was a facade, privately owned companies and their owners made huge profits on forced labor) or even thinking that the incorrect math of "nationalist" + "socialist" = "national socialist" is entry level humor at best.

I think it's a bit early to call the race.

I think the abstraction is a useful one - a general purpose classifier that does not need to be specifically trained: Unstructured signal in, structured judgement out - with a focus on speed and cost efficiency.

And since there is not yet a large body of benchmarks, I don't think we have sufficiently explored how to measure these things.

But since there's a market and some hype this will soon happen.

(And it's not like TypesafeAI has a real moat or invented something entirely new here ~ they just managed to put things into one coherent perspecive)

The difference is that the error messages contain instructions on how to resolve the issue.

Just flagging big functions will make the agent write small functions ~ but not necessarliy in a good way

(for example the agent might just cut `doOneThingAndTheOther` in half and call the second half `doOneThingAndTheOther2`)

> but I'd want to be able to setup a bunch of my own project specific code smells

That's what I did; I did not use the library I linked to ~ It served only as an inspiration.

Instead, I let the agents create custom lint rules (using eslint, pylint, ...) and add custom coaching error messages based on where I want to take my codebase.

Even if manual jobs were safe from being replaced by AI (for the sake of the argument, we're assuming that this is what's going to happen) these jobs would not be safe from the market pressure created by the displacement in other areas.

Here's a link to a video where the creator explains the idea

https://youtu.be/6AgndHSkHFI?t=238

(You could of course argue that you don't like the direction the rules push the agents in given in the example - some people don't prefer small functions everywhere - but that's not the point: The lint-hooks work in pushing the agent in the desired direction; If one desires something else they'd simply need different rules)

I think they even more so need deterministic feedback:

I tried an approach based on the following idea recently and it's amazing - Lint rules where the error messages contain an explanation on how to deal with the issue.

https://habit-hooks.com/

I'm using it to foster IOSP (integration operation segregation principle) for example.

openrig also seems to fit

Interesting. They're good in some ways and bad in others.

I was reminded of the amateur paintings by George W. Bush II.