djwide

---This is not written by AI --- Full Article Attached https://www.letters.senteguard.com/p/frontier-ai-firms-have-... https://youtu.be/uL3YDTw0bKU?si=LKvNwOrXkHoi4F74

TLDR - I am attempting to describe the open model firms' lack of moat for a more general audience. I am also framing frontier LLM tech as positive-externality tech like Linux and the Printing Press. Unfortunately for the Big Frontier CEOs this means they may not necessarily reap the benefits of the valuable models they helped to build.

What is a Positive Externality and What Are Some Examples?

A “positive externality” technology is one which generates societal benefit that exceeds the developer's personal financial return. For example:

— The Printing Press: While Johannes Gutenberg was the inventor of the printing press, he personally met financial ruin. The true economic value of the Press lay in the portability of language and the broader dissemination of ideas.

— Linux: Torvalds’ open-source operating system became the basis on which countless companies built multi-billion dollar businesses (e.g., Alphabet’s Android operating system).

An Analogy: Frontier vs. Open-Source as Rock Climbers

Generated by ChatGPT Imagine the top AI models as a group of rock climbers.

— The Lead Climbers (Frontier AI firms): Companies like Anthropic, OpenAI, xAI, and Alphabet are more capable at climbing and training the AI frontier. Because they are more skilled, the lead climbers establish the routes and place protection.

— The Followers (Open-source models): Firms like DeepSeek, Kimi, and Ollama come later. By training their models on already established frontier models in a process known as distillation, these firms approach the cutting edge faster than they otherwise would have given the work from the leaders.

Just as in rock climbing, sometimes lead climbers benefit from their fellow lead climbers—they use their routes or distill their models. Sometimes the followers lead in their own way as well, like when DeepSeek made massive efficiency improvements on model training in January 2025.

In this way, followers and their user base capture the benefits of the positive externalities generated by model training investments financed by frontier AI firms, their investors, and the American taxpayer.

...

Big Frontier’s Escape Plan

...

But Distillation is Unfair to the Frontier Firms? Aren’t the Open-Source Companies Stealing?

— The closed-source firms distill too.

— Frontier LLMs were and continue to be trained on the internet of human language —generally without compensation for the owners. Frontier firms are throwing stones from glass houses.

— OpenAI took funding and developed ChatGPT as a non-profit then transitioned into a “capped-profit” structure to secure the massive capital and computing power required for advanced AI development.

But they made the investment and they did the work. Don’t they deserve to reap the economic rewards?

— There is no guarantee that innovators will capture all, or even most, of the economic value they create. Linus Torvalds did not. Johannes Gutenberg did not either.

— Furthermore, these investments were funded not only from private capital but also by the American taxpayer under tenuous claims of imminent AGI, catastrophic AI risk, and an existential threat from geopolitical rivals.

I wrote a short essay arguing that the history of software development is fundamentally the history of moving up successive layers of abstraction.

Machine code gave way to assembly. Assembly gave way to compiled languages. Then operating systems, libraries, frameworks, and eventually Python. Each layer hid more implementation details and allowed programmers to express higher-level intent.

My argument is that LLM coding agents are simply the next step in that progression. We're no longer writing a programming language that resembles English—we're increasingly using English itself as the programming language.

The essay also discusses why higher abstraction historically increases human agency, why this makes AI a "normal technology" rather than a fundamentally unprecedented one, and what the next level of abstraction beyond natural language might look like.

I'd be interested in hearing where people think this analogy breaks down.

Linked youtube. Substack - https://www.letters.senteguard.com/p/the-next-rung-on-the-la...

Large Language Model (LLM)-enhanced authorship is accelerating at an extraordinary pace. Within academia, the share of papers crediting an LLM tool or model has grown exponentially since 2023. In software development, over half of all new code commits are now LLM-assisted. Largely due to LLM assistance, the rate of knowledge production has never been higher. The intelligence explosion will not be constrained by the limits of LLM capability, but by our cultural norms around attribution and by linguistic gatekeeping. Although intended to control the quality of academic work, traditional ideas of authorship within many disciplines may instead act as a buffer, diminishing the potential for human knowledge growth.

The accelerating capability of LLM systems to generate scholarly text highlights a longstanding tension within academia: the dependence on clearly identifiable human authorship as a basis for credibility. Universities and journals currently restrict LLM co-authorship, citing questions of accountability, transparency, and research ethics. These concerns are grounded in the principle that scholarly claims must be traceable to a responsible agent who can defend the work.

Legacy Attitudes Towards Attribution

Recent public discussions surrounding citation and attribution practices across academia have demonstrated that authorship norms have always involved collaboration, borrowing, and iterative drafting to varying degrees. Committee-produced writing, multi-author workflows, and the role of research assistants and editorial staff have long contributed to the final scholarly voice. The result is paradoxical: LLMs can make knowledge creation faster and clearer than ever, yet systems designed to ensure trust and credit are slowing its publication.

This conflict has played out very differently in software engineering. There, authorship is secondary to utility. Copying, pasting, and reusing existing code is not simply tolerated, it is the norm. Attribution norms are weaker not because developers lack ethics, but because their incentives are aligned around functionality. This norm makes software uniquely suited to rapid LLM integration, because LLM code assistants are a continuation of a long-standing culture of reuse. GitHub Copilot, for instance, builds on decades of norms around forking, patching, and sharing code with minimal concern for original authorship. As a result, software R&D will outpace other disciplines due to relaxed provenance norms.

In 1997, Garry Kasparov became the first world chess champion to lose a match to a computer. The machine, Deep Blue, used brute-force computation combined with heuristic evaluation in what was an early instance of machine learning. No human has defeated a cutting-edge chess engine since. However, even as humans lost their dominance in pure play, they have been successful against those same machines when playing in a human-machine pair. Competing alongside machines in a style known as cyborg chess, they routinely outperform both human grandmasters and standalone AI systems. This model offers a lesson for other domains of knowledge. The scholars of the future may become “cyborg scholars.” Their strength will not lie in generating ideas faster than machines, but in discerning which of those ideas are worth pursuing.

Continued at link or Substack / Youtube below.

https://www.letters.senteguard.com/p/cyborg-scholars https://youtu.be/c7DdLtGSux0

I'm saying there's something structurally different form autonomous systems generally and from an LLM corpus which has all of the information in one place and at least in theory extractable by one user.

I point that out a little bit when I refer to agencies being discouraged from sharing information. The CIA may be worried about losing HUMINT data to the NSA for example. You may be referring to them worrying about compartmentalizing the information away from the president as well which you are right happens to some extent now but shouldn't 'in theory'. Maybe it's a don't ask don't tell. I think Cheney blew the cover of an intel asset though.

Can anyone tell me why the comment gets downvoted. The article is past character count - I have to link.

Thanks for the comment. Interesting to think about but I am also skeptical of who will be doing the "collecting" and "synthesizing". Both tasks are potentially loaded with political bias. Perhaps it's better than our current system though.

There's not a direct tie to what I'm trying to sell admittedly. I just thought it was a worthwhile topic of discussion - it doesn't need to be politically divisive and I might as well post it on my company site.

I don't think there are easy answers to the questions I am posing and any engineering solution would fall short. Thanks for reading.

AlphaGo was trained on old games and then presented with a game it had never seen before. It came up with a move that a human would not have played.

Either way, not sure protectionism and siphoning money to frontier model owners will help us.

But also by that argument they would have beaten us to frontier model tech as well. Their education system appeared better than ours 20 years ago. We could have a bigger and broader conversation comparing the two systems and China's has a lot of flaws

It’s a call to patriotism. China versus America. “Who will you back?” This has become a common plea from the Silicon Valley elite over the last six months. I heard the move up close at the Harvard Kennedy School, where a visiting Eric Schmidt warned that AI may soon cross into autonomous self-improvement, argued that someone will need to “raise their hand” and impose limits, and then pivoted into the geopolitical register, contrasting American and Chinese trajectories and urging policy and funding choices aligned with “American values.” Others have also made versions of this argument in different forums. Tarun Chhabra, head of national security policy at Anthropic, has made a similar argument, urging an “American stack” and treating model governance as a geopolitical contest. Putting aside the awkwardness of nationalist messaging coming from the Bay Area’s long-time borderless “global citizens,” the incentives are not hard to see. If you can frame the open vs closed models debate as a national security referendum, you can frame restrictive rules as patriotism and you can frame “responsible control” as synonymous with dominance by a small circle of incumbent providers.

The posture makes sense once you consider two facts. One: industries which may live and die on capricious regulatory rule making must make their case to those with their hands on the levers of power. In 2026 America, those hands are professed patriotic Republicans. Two: Big Frontier LLM is losing the tech battle, or at least losing the easy assumption that America’s lead is automatic and permanent. They are on their back foot so they must frame the open vs closed model debate wrongfully as a fight between America and China. America cannot afford to lose a battle to China and by extension Anthropic, OpenAI and Alphabet cannot afford to lose to their competition.

Yet there is nothing inherently Chinese about open models and nothing inherently American about closed models. If anything, it is the opposite. Open models are decentralized, inspectable, forkable, and difficult to monopolize. That aligns with an American instinct to diffuse power, prefer competition over permission, and distrust single points of control. Closed models concentrate capability behind a small number of gatekeepers, wrapped in secrecy, and sustained by privileged access to regulators. That logic is far closer to centralized control than to open competition. The real fault line is not America versus China. It is democratic diffusion versus unnatural scarcity, and good tech versus bad tech.

Full article linked.

In 2000, President Bill Clinton famously looked at Beijing’s early internet controls and quipped: “Good luck. That’s sort of like trying to nail Jell-O to the wall.”

So far he’s been proven wrong. The CCP didn’t just contain the internet; it has effectively used the internet as a tool to entrench its control by building a system that fuses chokepoints, platform governance, and punitive enforcement into something like a sovereign information utility. That said, the jury is still out, and Clinton may still be vindicated.

On the one hand, LLMs can be understood as a natural outgrowth of Clinton’s (and Gore’s) internet but it can also be seen as its next evolution. By amplifying individual autonomy, LLMs present significant opportunities for economic growth but in pursuing growth they will also amplify individual agency. Therefore, the Party faces a quandary: pursue a strategy of economic growth and risk an erosion of Party authority or crack down and risk being left behind in the technology of the future.

Full article linked.

What do they (or you) have to say about the Lee Sedol AlphaGo move 78. It seems like that was "new knowledge." Are games just iterable and the real world idea space not? I am playing with these ideas a little.

With LLMs the synthesis cycles could happen at a much higher frequency. Decades condensed to weeks or days?

I imagine possible buffers on that conjecture synthesis being epxerimentation and acceptance by the scientific community. AIs can come up with new ideas every day but Nature won't publish those ideas for years.