Highlights of the Week
AI 101: What Is a Token (And Why It Runs AI)?
https://www.turingpost.com/p/token
English is often tokenized into words and subword pieces, because spaces clearly separate words and longer terms can be broken into reusable chunks. Chinese works differently: words are not separated by spaces, and single characters often already carry meaning, so tokenization tends to stay closer to the character level. That is one reason the same sentence can produce a very different token count in English and Chinese.
I’d never really thought about this before but it makes sense.
Cybersecurity Looks Like Proof of Work Now
https://simonwillison.net/2026/Apr/14/cybersecurity-proof-of-work/
If Mythos continues to find exploits so long as you keep throwing money at it, security is reduced to a brutally simple equation: to harden a system you need to spend more tokens discovering exploits than attackers will spend exploiting them. An interesting result of this is that open source libraries become more valuable, since the tokens spent securing them can be shared across all of their users. This directly counters the idea that the low cost of vibe-coding up a replacement for an open source library makes those open source projects less attractive.
There’s conflicting opinions on this but directionally it makes sense that more mature systems will be more battle tested for bugs and vulnerabilities - sort of like how it is at the moment then just a bit more turbocharged by AI.
Good and Bad Harness Engineering
https://danielmiessler.com/blog/good-and-bad-harness-engineering
Bitter Lesson Engineering comes from Richard Sutton’s “Bitter Lesson” essay, and it means ensuring that you’re not trying to outsmart your own AI. It means not trying to micromanage how your AI does things, but rather specifying what you want done. Plainly stated:
- Bad Harness Engineering is a whole bunch of prescriptive instructions on exactly how to do things. First copy this file, then load this, then do this, then do that. Etc.
- Good Harness Engineering is about providing tons of context about who you are, what you’re about, what you’re working on, what you’re trying to accomplish, and what good (and bad) look like to you. I’m an engineer focused on personal productivity, I like simple designs with lots of whitespace and great typography, here are my previous projects, here are some tools you can use, etc.
Putting into words actionable advice on how to use these new systems. It is partly why it always felt odd to me to go super deep into the optimisations around some AI models and tools that people find when unless you actually need that high level of performance you can just work around it for now and wait a few months for the next model which probably won’t have the same issues.
China Shock 2.0: The Flood of High-Tech Goods That Will Change the World
https://www.ft.com/content/7d51a630-a3de-4cc7-9f5f-0f3e7f0d305a
Alarmingly for governments across Europe, Asia and elsewhere, the political and economic factors fuelling the rise in China’s trade surplus are intensifying. The country’s protracted property slump and weak social safety net have curbed consumer spending, resulting in zero inflation last year and an increasing reliance on external demand to prop up growth.
This shock, echoing the first one, is a result of the structure of the Chinese economy, not simply a policy decision. Meaning that the Chinese leadership doesn’t really control it either as much as many expect they do so fixing it is going to be complicated without enormous pain for their economy which is unlikely to happen from within.
As Chinese factories rushed into solar, production capacity skyrocketed. The country has the ability to manufacture 1,200GW of solar panels annually, roughly double the 647GW installed worldwide last year, according to the China Photovoltaic Industry Association and energy think-tank Ember.
You Slipped Up. Here’s How to Get Back on Track
https://ryanholiday.net/you-slipped-up-heres-how-to-get-back-on-track/
Do hard things. Making a life change, adopting new habits, doing anything challenging requires courage. As I write about in Courage is Calling: Fortune Favors The Brave, we can’t just hope to be brave when it counts. Courage has to be cultivated. To do the big things that scare you, start with smaller things—start with developing the ability to push yourself to do stuff you’re reluctant to do. To be able to endure the cold reception of a bold idea, start with enduring a cold shower. To be able to step forward when the stakes are high, regularly do that when the stakes are low. To be able to embrace the discomfort of a major life change, accustom yourself to minor discomforts. We treat the body rigorously, Seneca said, so that it may not be disobedient to the mind. We push ourselves in little ways so the big ways stop seeming quite so big, quite so out of character.
Just make a little progress every day. For a long time, my writing habit was all-or-nothing—either I wrote a lot of words or I didn’t. Over time, I’ve lowered the stakes: now the question is simply, “Did I make a positive contribution to my writing today?” Sometimes that means writing, sometimes editing, adding, deleting. Sometimes I’m home and it’s in my office, sometimes I’m on the road and it’s on a plane or in a hotel room. Sometimes it’s a big contribution, sometimes it’s a little contribution. “Well-being is realized by small steps,” Zeno would say looking back on his life, “but is truly no small thing.” Focus on that—just making a little progress each day.
Most self help advice is best avoided I think, but a little can help so here’s my dose for the time being.
How NASA Built Artemis II’s Fault-Tolerant Computer
https://cacm.acm.org/news/how-nasa-built-artemis-iis-fault-tolerant-computer
Riley emphasized that while fail-silent logic is critical, it must be paired with active monitoring to avoid catastrophic gaps. “If a software component fails silently, the failure may go undetected unless monitored by another component or watchdog timer,” he said. For mission assurance, he said, error detection and recovery mechanisms must be explicitly designed and correlated across multiple layers of the codebase to ensure consistent behavior.
While the four-FCM primary system is robust, NASA must still account for common mode failures—software bugs or catastrophic events that could theoretically impact all primary channels simultaneously. To mitigate this, Orion carries a completely independent Backup Flight Software (BFS) system. This is a prime example of dissimilar redundancy. It is implemented on different hardware, runs a different operating system, and utilizes independently developed, simplified flight software.
Proper rocket science I guess. Or rocket software. Either way there’s likely things to learn from this, especially the monitoring one. Interesting that the backup one probably required enormous engineering just to do so it can be there in case the main one fails for whatever reason.
“Conviction Collapse” and the End of Software as We Know It
https://www.oreilly.com/radar/conviction-collapse-and-the-end-of-software-as-we-know-it/
He’s genuinely excited. But he described what he’s feeling as “conviction collapse.” As he put it, in the old world, you raise money, and nine months later you come back with a product. In that intervening time, you’ve talked to hundreds of customers. You’ve honed your worldview, and you’ve had time to build and defend your conviction. Now? “You invest in my company today, on Thursday I’m going to come with the same amount of stuff that would have come with nine months in the prior times. It’s just so fast. And so you don’t have the time to fall in love the same way. You just don’t have the time to enjoy and define and defend your conviction around your product.” That’s an eye-opening insight. Quintessential Harper. The result is that they build an entire product, complete with landing pages, show it to someone, get feedback, and then just build another entire product. Harper said, “Every time we hit a wall, we are like, ’Okay, what do we get from that?’ And then we just roll that learning into the next iteration.”
Everything has sped up and that may not necessarily be a good thing. There’s plenty we learn from spending time thinking about a problem that doesn’t happen now if AI does it all. If there’s no iteration there’s no learning, no feedback.
5 Years of Lessons From Running My Own Bookstore
https://ryanholiday.net/5-years-of-lessons-from-running-my-own-bookstore/
Doing interesting things usually pays off. When I was starting out as a writer, an author gave me a piece of advice I’ve never forgotten: If you want to be a great writer, go live an interesting life. He was right. Great art is fueled by great experiences—or, if not “great” experiences, at least interesting ones. That was in the back of my mind with the bookstore. Even if it failed, I knew the experience of trying to open a small business in rural Texas during a pandemic would be filled with stories. And it has been. I’ve drawn on it constantly—in my writing, my talks, in conversations with people on the podcast. So when you have the choice between the safe, boring path and the interesting one, take the interesting one. It always pays off.
More reason to do more interesting things and seek out new stuff.
The Second Wave of the API-first Economy
https://brandur.org/second-wave-api-first
Fifteen years ago, us API maximalists thought that APIs were going to eat the world, ushering in a new paradigm of interoperability that would vastly expand our capabilities as users, and even change the world for the better. What we got instead was an API winter. As useful as APIs were in some situations, that usefulness was outweighed by concerns around revenue, privacy, and abuse. But as scary of a thought as it was that this might be the end, it wasn’t. We’re at the beginning of a new spring of APIs that’ll appear to support use by agents acting on behalf of people. As this mode of operation gets more popular, expect the availability of an API to be a competitive edge that differentiates a service from its competitors. The result will be a global proliferation of APIs and expanding product capability like never before seen.
What is old is new again I guess. APIs might be what is described or might be just a stepping stone. Hard to say for now what might happen
Apple at 50: How Asia Fuelled Its Rise to the Top
https://www.ft.com/content/95fda471-0694-4445-8222-30e09b03887b
This tactic was repeated again and again: Ive’s industrial design group would find a Japanese supplier and talk up design, the significance of their work, and promise lots of business. But when operations, the division tasked with scaling at minimal cost, would inherit the relationship, they often found the Japanese to be expensive and risk-averse. Result: Apple would copy the craftsmanship, reverse-engineer it, and scale it elsewhere.
China and Taiwan, this person says, were great at building “good-enough machines”, whereas Japanese vendors often made machines that were “too good” — in that they couldn’t scale quickly enough to please Apple’s volume-focused operations team. Some former SMEs admit feeling bad about the practice. “I felt horrible about it,” says one. “We’d screw over the original IP owner. I did this over and over.”
In other words Apple did not merely outsource manufacturing to China, it played a starring role in building up its industrial capacity. The contrast with Japan is stark. Over the past quarter century Apple has extracted critical knowhow from Japan’s best minds, turning the country’s famous electronics brands, like Sharp and Sony, into invisible subcontractors inside Apple products. In China, invisible subcontractors, such as BYD, scaled the value chain and evolved into China’s world-leading brands.
Ruthless. I guess that is what you need to be, but now will we see someone out-Apple Apple at this. Chinese companies are already competing beyond many in the west. And what happens now if the craftspeople are no longer around, where does that innovation and refinement come from? Ideas come from somewhere and if everyone is copying everyone then where’s the originality coming from?
The Middle Loop
http://annievella.com/posts/the-middle-loop/
It occurred to me that we often refer to “loops” in software development - the inner loop and the outer loop: • The inner loop is where the craft lives: write code, build, run, test, debug. Tight, fast, local. This is what TDD and better IDEs optimised. • The outer loop is the broader cycle: commit, code review, CI, deploy, monitor, feedback. This is what DevOps and CI/CD optimised. What if supervisory engineering work lives in a new loop between these two loops? AI is increasingly automating the inner loop - the code generation, the build-test cycle, the debugging. But someone still has to direct that work, evaluate the output, and correct what’s wrong. That feels like a new loop, the middle loop, a layer where engineers supervise AI doing what they used to do by hand.
Software engineering is changing, it is not what it once was. We all need to learn and adapt to what it is going to be. Maybe this middle loop idea has something in it, or it is just a transition phase to the next stage.
Finding Comfort in the Uncertainty
https://annievella.com/posts/finding-comfort-in-the-uncertainty/
At the close, when the organisers offered the mic to anyone who wanted to share a takeaway, I took my chance. I challenged the group - all of us with platforms and voices - to help the current and next generation of engineers find their way. Not by reassuring them that everything will be fine, but by getting better at describing the new problems they get to solve now that writing code isn’t the main one anymore. There are many engineers out there who were hired to write code and are now being told to outsource that to AI. They deserve more than platitudes. They deserve us doing the hard work of defining what comes next.
Across security, governance, and adoption discussions, platforms kept emerging as the thing that makes everything else possible. The pattern: create safe defaults (the ‘pit of success’), embed guardrails, and give teams a fast but safe path. The argument isn’t for more controls. It’s for better defaults - so that speed and safety aren’t in tension. One group observed that the things that make agents effective (clear standards, good documentation, well-structured code) are the same things that made developers effective. We spent years advocating for Developer Experience and often struggled to get investment. Now organisations are keen to invest in the same things - but for agents. Is ‘Agent Experience’ the new DevEx? And if so, might humans quietly benefit too?
Agent experience is developer experience I think for the most part, only for now there’s more incentive by leadership to work for agents and not developers.
In at least four separate conversations - spanning agentic operating systems, production operations, enterprise adoption, and hallway chats - people independently arrived at the same idea: we need a complete, verifiable record of everything agents do. Every step taken, every change made, with the ability to roll back. The framing varied. Some described it as a work ledger analogous to a financial blockchain. Others as an audit trail for verification and accountability. Others as a unified log aggregating across infrastructure, software, and network layers. But the core insight was the same: if we’re going to trust agents with real work, we need a ledger that tells us exactly what they did and why. When this many people reach the same conclusion independently, it’s worth paying attention to.
GitHub competitors are going to start popping up. Nobody knows yet the right way yet.
The secret behind Japans railways
https://www.worksinprogress.news/p/the-secret-behind-japans-railways
Japanese roads are expected to be self-financing. Motorways are run by self-contained public cooperatives, very similar to the statutory authorities that ran English roads and canals between 1660 and the late 1800s, and funded by tolls on their users. Vehicle registration taxes, which are allocated to localities for road construction and maintenance, are worth three percent of the Japanese government budget. These measures, adopted in the 1950s, were not intended to suppress car use – the point was to fund a massive road expansion – but they have forced private vehicles to internalize many of their hidden costs.
Japan is one of the only countries to have privatized parking. In Europe and North America, vast quantities of parking space is socialized: municipalities own the streets and allow people to park on them at low or zero cost.
Expecting cars to pay their own way is one reason behind it. It leads to success stories like this when done right:
In 1953 the company decided to build the Den’en Toshi Line, or Garden City Line, to serve a rural area southwest of Tokyo. This would be enabled by a series of land readjustment projects collectively among the largest in Japanese history. Over 30 years, 3,100 hectares were covered, of which only 36 percent was devoted to residential and commercial development, with 20 percent for forest and parks, 17 percent for roads, and much of the rest for watercourses. The population of the land readjustment zone would rise from 42,000 in 1954 to over 500,000 in 2003.
The Age of the Amplifier
https://www.construction-physics.com/p/the-age-of-the-amplifier
A few months later Townes assigned Jim Gordon, one of his graduate students, the task of building the device, assisted by another graduate student, Herb Zeiger. Over the next several years Gordon worked diligently to realize the ambitious concept. Many physicists considered the idea unpromising: at one point, several years into the project, the current and former heads of Townes’s department, Isidor Rabi and Polykarp Kusch, exasperatedly stated to Townes that “you should stop the work you are doing. It isn’t going to work. You know it’s not going to work. We know it’s not going to work. You’re wasting money. Just stop!” But finally in 1954, a few months after Kusch insisted the device wouldn’t work, Gordon succeeded, demonstrating both amplification and oscillation in a collection of stimulated ammonia molecules. They called their device the maser, an acronym describing the device’s Microwave Amplification by Stimulated Emission of Radiation.
The story of science - this story is repeated over and over in many different fields.
Within just two years of Maiman’s demonstration, a laser was used for eye surgery. By 1968, the US Air Force was dropping laser-guided bombs in combat. By 1971, Xerox PARC researchers built the first laser printer, and in 1974 the first laser-based barcode scanners were being installed. In 1980, the first commercial fiber optics lines using semiconductor lasers were made available. It’s the lasers ability to generate coherent light — light all of the same frequency, and in the same phase — that made it possible to continue to advance up the electromagnetic spectrum, and use optical wavelength electromagnetic radiation for communication.
Things can move fast when the needs are there for it to do so.