'(:JONES)
1.4K posts

'(:JONES)
@infix_fun
he/him, software engineer. clojure. ai tooling.
Rio de Janeiro, Brazil Katılım Mayıs 2024
343 Takip Edilen27 Takipçiler

I hate to admit it but Opus in Cursor has been performing better than GPT over the past few days.
teej dv 🔭@teej_dv
5.5 just edited a node_module file directly.... what?!
English
'(:JONES) retweetledi
'(:JONES) retweetledi
'(:JONES) retweetledi

@flowstated Thanks for all the latest deliveries and bug fixes. Could you guys optimize multi task to always span sub agents and leave the main agent free? This has been a recurring issue as an user. Look forward to using it more.
English

try /multitask in cursor today

erik@flowstated
cursor now has multitask mode stay in flow with one thread + let the agents cook
English

'(:JONES) retweetledi

what is it about physics and math that make genius mfs entirely lose their mind
Kekius Maximus@Kekius_Sage
Nobel Prize physicist Frank Wilczek says matter, energy, and even reality itself may ultimately emerge from information.
English

@android_poet @csinco Some people care about the work they do you know
English

@csinco I don’t know why people love discussing useless stuff. Just do the work assigned to you, go home enjoy life, and spend time with your family.
English

@championswimmer Unfortunately, some people think that mathematics is equal to thinking. Worse, they think if you don't formally define something there's no worth or not work done. It's really sad and it goes to show how many people in STEM are out of touch with humanities
English
'(:JONES) retweetledi

This is one of the smartest people in the world, without doubt. And I feel sad about how tunnel visioned the AI folks have become + on top of that they feel they have all the best ideas in the world because they are developing the things that's setting the economic agenda of the world.
But is this posture below even truly defensible? Have you never watched a scene in a movie and truly felt in awe of either the emotion it evoked or the spectacle that unfolded. Have you never wondered how the director came up with the "idea" of that scene? Because in their mind they had the whole thing before it even existed.
You think they express their ideas in mathematics and code?
You think people who paint, who produce music, who write stories have no ideas? Or you believe all of their ideas can be mathematically expressed?
How sad.
English

@Duodecahedron12 @championswimmer Are you a musician by any chance? I'm sure you're not. Do you think musicians are out there expressing music like we write an equation? The fact that math is everywhere doesn't mean formal notation is on people's mind. I mean you'd have to be completely out of touch to think so.
English

the language of mathematics and code is in everything, including music.
So if you are a musician, you are expressing mathematics.
If you are studying something deeply, you can see the mathematics in that thing.
For example, my daughter is a musician and she studies the mathematics of music through music theory.
This is aligned with the Platonic world view.
Its not a personal attack.
Why are you so anti-intellectual?
English
'(:JONES) retweetledi

I'm such a LISP person, ngl. I actually never realized it as clearly before.
Saagar Jha@_saagarjha
I will not be taking questions at this time.
English
'(:JONES) retweetledi

@Dr_Gingerballs @iam_multiman Perfect is a useless expec6. We need working code and code that satisfies the requirement. Making it well suited is the engineer's job.
English

@iam_multiman Coding is highly degenerate. Many solutions exist for a set of parameters, and most of the solutions are not well suited to scale over time in a complex code base.
There is no perfect code, so there is no perfect coding agent.
English

An analogy for why I believe current AI coding agents will not survive in a meaningful way long term.
First, what is an AI coding agent? It’s a large language model trained on all of the open source code available on the Internet, attached to some sort of loop. You ask it to create a program that has some functions, and provide it with details about how it must operate. The output process then follows:
1. The LLM outputs code as a guess.
2. The looping tool evaluates the code in some way based on stated functional requirements.
3. If the code does not pass, query the LLM to make another guess.
4. Continue until an exit condition is satisfied or you run out of compute.
It may not seem like it, but this is just iterative, fuzzy search optimization, just over written words.
The utility of the system depends on the quality of the guesses, the evaluation mechanism, and the optimization strategy.
The quality of the guesses depends on the quality of the training dataset. Does the training dataset contain the code snippets needed to make your request?
For simple and common requests, the answer is yes. If you just need an efficient sort routine in a language you aren’t fluent in, you can get the model to make one for you and it might save you 10 minutes. Not insane speed up but definitely compounds over time. Here the coder knows what they want exists, knows how the sort algorithm is supposed to work, and just needs one whipped up in a new language they are building in. The expert saves some time. Integration into the codebase is still done by the human. It’s basically fancy autocomplete.
For more complex requests, such as multi function routines which require a large amount of architectural design, the agents start to fall apart. This is because the likelihood that someone has built exactly what you wanted goes down quickly as the size of what you want increases.
Here enters the loop. The agent producers hope that your request is similar enough to a range of existing code that they can guess a workable version by interpolating (and sometimes extrapolating) between solutions. So they make a guess with some randomness applied, evaluate, and modify the guess based on the results.
Anyone who has done iterative optimization can identify a lot of the issues that occur in these systems.
You might get stuck in a suboptimal state, where all the next guesses are worse than the current guess, even though the current guess isn’t an acceptable solution. The output seems like it’s almost there but not quite. The user then keeps requesting more iterations, hoping to go from 90% to 100% that never comes.
There may also be degeneracies in the sample space, and you might get something that passes the criteria but is sloppy, nonsensical, or ridden with unnecessary bloat under the hood. Like a root finder that just won’t find the root you are looking for.
And so in the course of writing, say 1000 lines of code, the agent has actually written 1M lines of code, iteratively generating and praying it can pass off as acceptable. The user never sees most of this, just told the system is “thinking.” When all is said and done, that 1000 lines of code required the generation of millions of lines of code, mostly thrown out.
Now to get to the analogy. Think of the agent as a bricklayer and you have asked for a brick wall. You specify color, pattern, accents, etc.
But the bricklayer isn’t very skilled, and decides to lay bricks stochastically. First, he evaluates each brick after placement. Thickness of seams, alignment, angle, etc. if it is wrong, he breaks it out and tries again. For every brick in the wall, he lays 100 bricks and wastes 99.
Then he decides to go faster, only evaluating every 10 ft of wall. If there are more than 10% errors, he destroys it and rebuilds. For every 10 ft he lays 1000’s of ft.
English

@mattpocockuk Don't you think claude code should expose those flows in a better way than just calling skills?
English

A flow I just tried and LOVED:
1. /grill-with-docs, talking about a new bit of UI
2. Asks me a question I can't answer unless I prototype
3. /prototype
4. Iterate on the prototype, burning tokens freely until we get a good spot
5. /rewind to the question, and select 'summarize' (Claude Code feature), saying 'summarize what we learned from prototyping'
6. Continue the grilling session, retaining the prototype
Smoooooooth
English
'(:JONES) retweetledi

'(:JONES) retweetledi












