The Four Levels of Agentic Software Development in the Enterprise
Agentic development from a platform engineering point of view.
From my reading, I also find “slot,” “harness,” and “guardrails” to be annoying.
Will we be able to notice these differences in the future?
I think Inference Engineering will be the future (or is it already here everywhere?). There is a lot of work in this area: managing memory, both short-term and long-term (can the agent remember what I did previously?), agent performance evaluation (does the agent finish the task as requested?), etc.
But he also seems frustrated with self-revolution’s results thus far. He has complained, with unusual frankness, that “if I do not write comments on reports, no work gets done.” Among the millions who staff the party-state, he sees too many officials with “inadequate understanding,” “old-fashioned concepts,” and “insufficient ability.”
If true, it’s pretty remarkable.
The wager is that the party does not need elections or rule of law to remain effective. Internal discipline and self-correction, Xi believes, can generate enough accountability, legitimacy, and success to sustain its rule.
现代人体能量危机，正在拖慢我们的思考 (Chinese)
Actually, any exercise that keeps your heart rate in Zone 2 can significantly improve mitochondria production.
The Water Use of Data Center Workloads
Water usage in data centers is not as bad as people might think.
Why Japanese Companies Do So Many Different Things
And this means that Japanese companies strive to avoid financial pressure from outsiders. Relationships with suppliers are longstanding and entrenched: many Japanese companies have been working with the same suppliers for 50 years or longer. Outside investors seeking to interfere in this happy picture will find few avenues for influence. A standard Japanese firm’s board of directors is composed almost exclusively of the firm’s own senior managers; a large fraction of the firm’s equity is held not by outside investors but cross-held by other Japanese firms; and most of the firm’s financing comes from a single “main bank” that provides loans and monitors performance.
And as a result, Japanese companies don’t really try too hard to return profits to shareholders. Earnings are mostly reinvested, and investor dividends are kept low. For a long time, Japanese firms would spend as much entertaining the managers of other firms as they would on dividends to shareholders.
But in order to understand each other’s jobs, workers cannot be specialized: they have to rotate across different workplace functions to the point where they’re familiar with much of the plant’s operations. In order to rotate across different workplace functions, they need broad training; and it makes no sense to train them broadly if you don’t keep them for a very long time. And if you have generalist workers who are around for a long time, you can’t reward them based on how they do in one role, because then they’d have no desire to leave that role for another role where they might do worse. Instead you have to pay them based on company performance, and promote them based on seniority. And you also have to give them an ironclad commitment not to fire them if economic conditions worsen: if they can get laid off at any moment, why would they invest years of effort in learning all the idiosyncratic things that your firm does.
And that basic impulse toward survival is why Japanese companies are so insistent on diversification. If you’ve made a commitment to keep people employed for life, then you need to create jobs for them if their current jobs stop making sense.
And this system, as it turned out, was really good at particular things. Aoki’s key insight was that the J-mode had a comparative advantage in environments of moderate volatility: situations where conditions changed frequently enough that rigid central plans would be outdated before they were executed, but not so radically that only top-down strategic intervention could cope. In an environment of stable, predictable demand, the H-firm did fine; in an environment of extreme disruption, where the whole product line had to be rethought, centralized authority was indispensable, and the H-firm also did fine. But in between—where the challenge was to make constant small adjustments in a changing but recognizable paradigm—the J-firm excelled.
I heard from a friend who worked at Fujifilm a long time ago that they were pivoting toward being a cosmetics company because film was dying.
While large Japanese companies do so many different things, many small companies (restaurants are a good example) only do one thing.