Sabitlenmiş Tweet
Robert Balicki (👀 @IsographLabs)
17.9K posts
Robert Balicki (👀 @IsographLabs)
@StatisticsFTW
@isographlabs framework author. Currently @Pinterest. Ex-React Data Team @Facebook. Co-organizer of #RustNYC. I like Rust, Relay, stats, GraphQL, React, JS
Manhattan, NY Katılım Nisan 2012
1.3K Takip Edilen1.8K Takipçiler
Robert Balicki (👀 @IsographLabs) retweetledi
new york's datacenter moratorium just froze a 500 megawatt project in genesee county that was designed to run on niagara falls hydropower. the state paused a computer powered by a waterfall to spend a year studying the environmental impact of computers
English
@naman34 So there's a fix everything button but we're not pressing it ;)
tbf Hochul is doing great work on energy production, so this is tweet is certainly a tad unfair to her
English
@StatisticsFTW TBF, energy use is a real issue. And governments are slow.
English
Robert Balicki (👀 @IsographLabs) retweetledi
Robert Balicki (👀 @IsographLabs) retweetledi
i've come to the conclusion that AI-generated prose is actually irredemable, and we've banned it from external comms at Archil
not only does is it very easy to spot, but it easily contaminates your thinking making it *really* hard to recapture any sense of authenticity
(once AI outputs an outline of talking points, it's harder to come up with better ones yourself)
we've been bad about this in the past, and so i'm doing a manual rewrite of the docs this week to give a more obviously human tone
Matt Silverlock 🐀@elithrar
just saw probably the worst (so far...) AI slop deck ahead of an ext. meeting today (they are presenting to us). it goes without saying: please do not do this.
English
The increase in content from my mutuals in my timeline is a clear improvement
English
@StatisticsFTW it's gonna get worse bc the ai illiterate are the most vulnerable to the ai slopaganda
English
"AI can and must be a force for good. At the same time, it could literally eliminate many entry and mid-level jobs that are the pathway to the middle class and beyond. If that occurs, we can have an economic and societal crisis beyond anything our generations have witnessed." -Hochul, March 19, 2026
Robert Balicki (👀 @IsographLabs)@StatisticsFTW
English
Robert Balicki (👀 @IsographLabs) retweetledi
My timeline was always fixed because I know how to click on the “not interested” button
English
barnum-circus.github.io to toot my own horn, I think a high-level language focused on type safety and making it easy to coordinate async work that is easily written by agents has a role: the agent can write a Barnum program, that program can then execute the larger task you could not one-shot
e.g. I use barnum to (among other things) read files and identify refactors to do
English
The era of niche languages is coming!
Aram Hăvărneanu@aramh
I am using LLMs with my own WIP language, which doesn't even have a definite syntax yet and dare I say it does a better job than with webslop. The lack of training data seems to be less of a problem than people think. Rather, the "better" the language is (by some metric), the better it performs.
English
@krismicinski @henrytdowling I would love to be able to make a statement and have it turn into an immutable law for the LLM. But inevitably even the most strident statements eventually get ignored. But not so for unit tests or types
English
@henrytdowling @StatisticsFTW AI is great but English language prompting to create apps will be remembered as an exciting time but ultimately crappy UX. Lots of issues with just raw steams of English prompts, since you do actually want something with a coherent semantics much of the time
English
@henrytdowling Niche languages can have advantages. For example, I'm excited about frameworks/languages that really, truly enable local reasoning w/something like a compiler that handles plumbing. Compare something like manually fetching and threading data vs relay/isograph
English
@StatisticsFTW dumb q but why are niche languages exciting at this point? like, are you writing code by hand?
English
Robert Balicki (👀 @IsographLabs) retweetledi
I am using LLMs with my own WIP language, which doesn't even have a definite syntax yet and dare I say it does a better job than with webslop.
The lack of training data seems to be less of a problem than people think. Rather, the "better" the language is (by some metric), the better it performs.
English
@kenwheeler It still did, but it at least didn't have to manually update the file... it had tooling to edit it (jq-style) and what amounted to unit tests
English
@kenwheeler I had an agent making configuration files for keyboard remapping that were 10s of thousands of lines long, and I built custom tooling to make it harder for it to fuck up
English
i’m actually curious, how many of you have actually written a tool in the literal sense for an agent
English
A case I’d like to see — these kind of preservation laws violate first amendment freedom of expression.
Matthew Zeitlin@MattZeitlin
you have no idea how many YIMBYs you've created with every landmarking
English
Robert Balicki (👀 @IsographLabs) retweetledi
You’re listening to Sisyphus FM, the home of non-stop rock.
English
Robert Balicki (👀 @IsographLabs) retweetledi
@halogen1048576 it just significantly degrades performance because what I *want* is an accurate factual assessment of a question, and what I get is a system that strongly asserts unqualified wrong information in ways where it’s easier to do the search myself or switch tools than correct it
English
@deepvaluebettor You really have to be vigilant
English
@StatisticsFTW after a few days of experimenting w pure vibecoding && not looking at any code , i ask for a graphviz to demonstrate the insane amount of accumulated tech debt + how fucking idiotic && chaotic the architecture had become
example 1:
English
Using LLMs to build presentations with lots of graphviz visualizations is a dream come true
English