arcsincosin

In the examples, there is discrepancy between the source, the diagram, the pieces diagram, and the 3D test each. As someone who is professionally and personally interested in grammars and textiles, I appreciate the fantasy and the vibe of the project but it does not seem to be a functioning demo of anything coherent.

This essay quickly superimposes the "throughput oriented" procedural shift of lean manufacturing onto the current LLM coding moment and uses our value language around manufacturing to tell a story about the near future of software. I find it well written and presenting a clearer hypothesis than other narratives about the LLM-replaces-junior-devs post-ZIRP moment.

I like the tone of SOLSTICE 5 by P Chadeisson as well: https://www.youtube.com/watch?v=cntb3wcZdTw

I'm building a simple web-based drawing app for home sewing patterns. It's a little bit CAD and a little bit sketching tool. Born of my desire for easier and 100x cheaper home drafting tools. The app is still quite early (adding Bezier tools today) and the core workflow is not functional: of draft, lay out, print, repeat, only draft is working. But one step at a time! https://github.com/twaldorf/flatland-client

My read–which may be wrong–is that much of the article is discussing applications where the end user is interacting with an interface that queries an LLM using baked-in prompts (in one case, a marketing content generation tool). These prompts are being written by the PM. The PM is not writing prompts LLMs to generate code, the PM is writing prompts which are hidden behind a web form or button or something in an interface, hence the prompts being part of the codebase. The author argues that when a PM is editing these prompts they are delivering an artifact that looks more like an engineer's artifact than a PM's artifact, traditionally.

This is wonderful. As part of a cryptography course in undergrad, I spent a fair number of hours adding/multiplying points on small curves by hand and eventually computationally, and seeing these animations represents the "feel" of that process better than any resource I have seen.

It seems that elliptic curves are used quite shrewdly in public key exchanges, used as a sort of off-the-shelf arithmetic system that renders some benefits (harder to solve the elliptic curve discrete log problem) and drawbacks (more expensive to compute, because of these point ops) compared to key exchanges relying on the integer discrete log problem. There is generally much energy spent on explaining the point operations but not much on explaining this context: this is an alternate arithmetic inserted into the Diffie Helman scheme that is secure and not so hard to work with.

The finite fields in application (e.g. in X25519) are so large compared to the toy examples that the technical details matter more for understanding the performance of the algorithms than the actual cryptographic method. Understanding (and convincing yourself of) the core behavior of the key exchange itself is perhaps best done with the number theory and group theory lens. There was a discussion here on HN a couple days ago which largely bashed the theoretical approach in favor of simply cargo culting the tools, which was appalling. I hope they all enjoy this link!

Anyhow, the application of elliptic curves to cryptography is clever and I would recommend everyone with an interest (and a little background in algebra) read Koblitz's paper[1] and Miller's earlier description [2] if they are looking for more context for OP's wonderful presentation.

1: Koblitz: Elliptic Curve Cryptosystems https://www.ams.org/journals/mcom/1987-48-177/S0025-5718-198...

2: Miller: Use of Elliptic Curves in Cryptography https://link.springer.com/chapter/10.1007/3-540-39799-X_31

I agree but it's worth noting that Beeple was selling JPGs for $5000+ in the mid 2010s as a commissioned artist in the music industry. So the NFTs have also brought some transparency to a successful digital artist along with the crazy valuations.

Yes, this exists! [1]

[1]: https://wildcards.world/

I'm having a hard time getting instruments to synchronize in the examples—over time all the sequences drift. Is this expected / known?

This is a good point: opting out of reality is a worse strategy for happiness on a longer timescale. However, reading the 24-hour news cycle and being aware of the real political landscape is not perfectly correlated; I'd argue it's 50% overlap at the upper bound and sub-ten % at the lower. Furthermore, reading the news cycle and being politically active in a way that affects your, your family's, and your "heirs'" wellbeing is probably almost negatively correlated.

Any political wins made by groups exploiting and dominating mainstream media (Trump) are explained by the fact that the barrier to entry in political action is so low—just getting a few people to vote is politically effective because almost everyone is just watching the news instead of doing anything real. Meanwhile Trump's team meets with money and accrues powerful stakeholders by annexing their agenda, building support in institutions and markets.

So ignoring political news and pursuing interfaces with actual nexuses of power is far more politically effective than the hundreds of people with twenty tabs of The Atlantic open and the WSJ draped across their lap like a religious shawl every morning. Go out, meet politically active people in your city, email representative staff, make friends with people with money and vested interests that are vulnerable to political changes.

There is a lot of well-aimed critique in this comment section and the piece could benefit from rigorous editing and scrapping of the examples which are very unfortunately chosen. I think the steel man for this piece is: 1.) Qualified generalists and charlatan generalists look the same if you aren't an expert in the fields in which they opine. 2.) Experts in one field using their native lens to make conclusions about another field is not a polymath practice but something closer to a hammer that sees only nails.

I think we, as software people, are susceptible to 2—software is successful in some cases because of this kind of cross contamination where it replaces the tools and methods of a separate field with its own, sometimes with great success (art, finance).

Decentraland without user-generated content? Club Penguin Island all over again?