groceryheist

WOW

Claude Code is rare product that is both beneficial and economically addictive, where its use increases demand for itself, at least in the supply / demand range for code we are accustomed to. It makes making software so much easier that Claude coding custom software becomes a solution to all sorts of past annoyances. Maintaining the software is easy enough thanks to Claude code.

This is correct. AI is a huge boon for open source, bespoke code, and end-user programming. It's death for business models that depend on proprietary code and products bloated with features only 5% of users use.

I'm thinking a lot about this currently as a recent convert (as of Opus 4.5). I think this post is on the right track, but like much of this discourse, it isn't really addressing how the technology will grow and the disciplines will adapt.

I'm by no means a doomer, but its obviously a huge change.

Generative coding models will never be 100% perfect. The speed of their convergence to acceptable solutions will decline in complex and novel systems, and at some point there will be diminishing returns to increasing investment in improving their performance.

The cost of software will fall precipitously and it seems unlikely that the increase in the value of programmers / engineers as they currently practice will offset the decline in the price in software. However, following the law of supply and demand, the supply and the amount of software produced will surely grow, and I think someone has to use the models to build software. I expect being trained in software engineering will be very helpful for making effective use of these tools, but such training may not sufficient for a person to succeed in the new labor market.

The scope of problem that a valuable engineer is expected to manage will grow enormously, requiring not only new skills in using generative coding/language models, but also in reasoning about the systems they help create. I anticipate growth in crossover PM / engineering roles. I guess that people who generalize across the stack and current sub-disciplines will thrive and valuable specialties and side-disciplines will include software architecture, electrical engineering, robotics, communication, and business management.

Some people will thrive in this new field, but it may be a difficult transition for many. I suspect that confusion about model capabilities and how to make the most of them and which people are doing valuable things will put a lot of friction and inefficiency into the transition time-frame.

Last thought, given how great models are at coding compared to general of knowledge, administrative, and bureaucratic work, I expect models are widely used to build systems that are supply shocks on such work. I don't think my argument above applies to such workers. I'm worried most about them.

2015

I dunno, I'm not a policymaker. What drawbacks do you see?

Only if visas can only be paid through earned income or returns on investments made with such. Otherwise you're mixing people who bring value by contributing labor with people who contribute capital. Both can be nice, but should we treat them the same or independently?

You can still require people to sustain employment in their field. Maybe companies can attest that a particular role classification requires a type of high-end talent. Auditing or otherwise verifying the attestation addresses the current allegations that H1-Bs are given for some jobs not requiring high-end talent.

The huge fee won't solve the cheap labor problem, only shift the equilibrium. The USA Tech job market faces increasing competition from Canada and Eastern and Southern European countries with lower wages but competitive talent better than available from generalist outsourcing. The new policy accelerates this trend as companies will seek to transplant workers from the USA into other countries. This is bad for American workers whose status as the geographic center of the organization declines.

In my view, the real problem with the H1-B program stems from the sponsorship system which ties each employee to a particular company and role. Unable to leave their position without threatening their residency, they are more willing to demand abuse (e.g., long working hours, poor leadership, subpar compensation) than the labor market requires.

An improvement to the program would make it easier for people to change job. Perhaps the government could permit highly skilled individuals to qualify personally for the visa so long as they sustain employment in their field.

Haha! I (with some assistance from stockfish) beat it on the first try.

I use vimium now, but I think with an emacs-based browser I would be better at using the advanced features.

This is so cool! I'm someone for whom emacs has steadily expanded its role in my computing life, but who will never adopt a text-based browser as a daily driver. Looking forward to the stable 4.0 release when I'll be prepared to use Nyxt and hope it can replace Firefox / Chromium as much as possible for me.

Two reasons:

1. Preprint servers create DOIs, making works better citable.

2. Preprint servers are archives, ensuring works remain accessible.

My blog website won't outlive me for long. What happened to geocities could also happen to medium.

Would be really fascinating to learn about how the most intensely engaged people use the chatbots.

Is it just me, or does this seem like big news?

Filecoin? Not sure if that's considered a "smart contract" since it's not eth-based.

If we're "lucky" (in an AI-optimist sense) we'll need the nuclear plants despite efficiency increases.

I fear this post's argument incorrectly assumes that the cap on insurance provides as 20% percent of revenue is the limiting factor on insurance company profits. In fact, it mentions cases where insurance company profits are below the cap! Their interest is not obviously to grow healthcare spending. So I'm not sure to what extent they truly endeavor to deny care so as to increase profit in the long run, and not just smooth a turbulent market.

This is the latest in a long line of attempts to make a non-centralized Wikipedia. A Wikipedia (not me) maintains a record

https://en.wikipedia.org/wiki/User%3AHaeB%2FTimeline_of_dist...

Nvidia is doing fine :), but it would be better for them if they were not limited. Hopefully, and probably, you are right and production is limiting.

The topline is that the meteoric growth has already occurred. 18% quarterly revenue growth is good, YoY growth was 15x that.

I like the recent work you've been doing Ed, but I think stylistically you can improve by seeing the truth in some of these criticisms. Your analysis can use a stronger systemic account of both the external cultural and market forces at work and even more so of the internal dynamics within large firms. It can also be more cutting by being concise --- avoid repeating yourself and non-substantive elaboration, especially in an emotional tone, which can be fatiguing. This will help you keep me as an audience member.

> First is that they can have agents check the cited source and verify whether the source backs up what's said to a reasonable degree.

This is a hard and tmk unsolved NLP/IR problem, and data access is an issue.

> Second opportunity is fitting in things only found in other language Wikis that either be incorporated into the english one or help introduce new articles.

This has been attempted via machine translation in the past, and it failed because you need native speakers to verify and correct the translations and this wasn't the sort of work that people were jumping to volunteer to do.

This is exactly right and is one (among many) reasons that reliance on cosine similarities in my field (computational social science) is so problematic. The curvature of the manifold must be accounted for in measuring distances. Other measures based on optimal transport are more theoretically sound, but are computationally expensive.

This review is enlightening in describing Graeber's break from conventional anarchism in the Dawn of Everything. Early Graeber tries to make the best possible case for critiques of power structures that characterize our modern world or underpin its legitimacy --- critiques derived from or harmonious with those in conventional anarchist thought. I disagree with the author's interpretation of The Dawn of Everything as evidence of Graeber shifting toward liberalism.

The Dawn of Everything is actually a significant contribution to anarchist thought (though it may be fair to call it post-anarchism) because anarchism struggles to theorize and build alternative institutions that can sustain and protect an anarchist society from domination by external powers. Graeber's earlier idea of prefigurative politics provides a partial solution because one can experiment with anarchistic institutions within a capitalistic / statist society.

The Dawn of Everything implicitly addresses the limitations of prefigurative politics, which are obvious in practice. Prefigurative institutions that are not short-lived or small-scale are rare and typically grow their own ideosyncratic power structures e.g., the tyranny of structurlessness or Wikipedia. Although not approaching a "state" such structures make anarchists uncomfortable especially when they reproduce hierarchies from the broader society.

The Dawn of Everything teaches us not to equate domination with the state or necessarily with hierarchies either. The posted review takes issue with how late Graeber rejects the concept of the state. But this rejection in no way lets the state off the hook. The Dawn of Everything is unequivocally critical of the state and seeks to understand how the state stabilized and persisted as a world-dominating organizational mode. It pursues this ambitious, (and I think unaccomplished) goal by describing a wide range of early societies and demonstrating that some were violently coercive without state-like institutions (e.g., north american slavers) and that others were peaceful and egalitarian but had institutions that other work associates with violent coercion or state emergence (e.g., cities or agriculture). Then, by analyzing the stability of these societies and how they worked it tries to piece together a theory of decomposed types of coercive power. As the reviewer points out, this decomposition isn't that theoretically satisfying. There could be other ways than the three types of power, and some of the arguments around this part of the book in particular seem to stretch the evidence.

That said, it is useful for anarchists to recognize that just because a power structure or hierarchy emerge within a political project or organization that it is a failure. Opposition to the state and domination more broadly doesn't require commitment to design principles like leaderlessness or flat organizing structures. The book's most important contribution is to show that human societies have already explored a vast design space of political institutions in our history.

Anarchism has always been a "liberal" philosophy --- indeed the most extreme form of liberalism. Any state-socialist or communist will say this. It opposes the state because the state extended the scale and reach of coercive structures like conscription, taxation, and private property far beyond their pre-modern limits. Yet the state is losing power to international governmental organizations on one hand and international corporations on another and so it is incredibly useful to think about ranges of better possible futures instead of doubling down on tried and tired commitments to ideological purity with a movement that was most significant and non-academic 130 years ago when the modern state was still contested in much of the world. Anarchism is stale but post-anarchism, like the Dawn of Everything, is essential.

Official USA inflation numbers ticked up slightly in December.

Yes. It is time for laid off seniors to take these exit packages and incubate the next generation. Be ready to draw financing when rates fall in Q3 of 2024.

Yes, of of the hard and interesting legal questions is if creating a possibility of such attacks constitutes a copyvio.

I agree that this sketch comes closer to working in practice than simple RLHF. In my earlier comment I was imagining bringing in some auxiliary data like you describe to detect plagarism and then using RL to teach the model not to do it.

Yes and the fact that the number of unique phrases grows so quickly with n is why the bloom filter needs to grow so that hashed n-grams don't collide.