Grief and Metacognition

Hello, A Newsletter by Arne Brasseur Sep 22

(View in browser)

Hello!

It’s been a little over a year since Claude Code came out, and agentic coding really went mainstream. Only a year! It’s been a wild ride. One piece I keep coming back to is this post by Brad Frost about Grief in the AI age. As software professionals we’ve had to rewire our understanding of what it is that we do, what our industry does. Our roles and job prospects are being challenged. Established wisdom is being discounted and disregarded. In the face of such upheaval, grief is a natural response. And so we find ourselves in denial, angry, bargaining (maybe with the right tools, with the right workflow, it’s not so bad, maybe I’ll have a little AI, as a treat), depressed (don’t take the black pill, though!), and finally in acceptance, coming to grips with the world as it is, and finding a working relationship with the present reality.

Naturally the think pieces are spanning that gamut, like this article, LLMs and performative productivity which crossed my feed this week (originally published in June). On the surface it’s a takedown of the notion that LLMs make you more productive, but what you really sense underneath it is the anger. The author seemed to have gone through some kind of awakening. I was feeling so good, so productive. But I’ve been deceived. It was all a lie!

They make some good points. Points that people have been making for well over a year. What strikes me at this finding out stage, as more of these experience reports come out, is how much of it has been predicted. For a certain type of experienced developer the downsides, the risks, the traps, were obvious from the start. Instead of being listened to, they were told to try the latest model.

I just came back from a week in the French Pyrenees. I did not bring a laptop, but I did bring two books, a biography of Manu Chao, and The Psychology of Software Teams. In the latter Dr. Cat Hicks talks about metacognition, about having the self-awareness and presence of mind to reflect on your problem-solving strategies.

Metacognitive awareness keeps you honest about your own abilities and productivity. It helps you to distinguish between busywork that feels productive, and focused work that actually moves the needle.

Much of what I do is simply asking the question "are we being smart about this?", and then trying to reflect deeply on that. On the state of the industry, on "best practices", and on what other choices we have at our disposal. There often is an "obvious" choice, the hegemonic framework, the architecture du jour, the methodology everyone is familiar with. And then there’s the approach that snugly fits your context and the problem you are trying to solve.

The IT industry is overflowing with self-deception. About which technologies are superior, which methodologies work best, and about the relationship between workers on the development floor, and "the business". Senior engineers with high metacognitive awareness are acutely aware of this. Navigating that gap between "industry standard" and reality is what their career is made of.

These are the people who could predict on day one the mess that LLMs would cause in our teams and codebases. Not only because they understand the limitations of the technology, but also because they knew this would be rolled out uncritically across the industry.

From "LLMs and performative productivity"

For any purpose where measuring productivity matters, you can’t just look at speed or volume. These are short-term metrics, and good work is often done slowly, and in small increments, in order to last long-term.

But it seems like we don’t care anymore.

We, as an industry, never did. It’s an uncomfortable truth for a lot of developers who, as much as anyone, want to feel a sense of pride and purpose in their work. If you work in an organisation that still invests in quality and long-term impact, then you are one of the lucky ones. The norm is chasing short-term productivity, churn out features and move on. This is not new to AI. It’s been true since at least the SaaS boom of the 2010s. It’s why the promise of agentic coding is so alluring, and why it’s so hard for people to really grok the downsides. They are exactly where our collective blind spots are.

The fifth stage of grief is Acceptance, and Commitment. Acceptance means you are not in denial about reality. You don’t have to love it, but you should be clear-eyed about it. Reality is that AI-assisted coding is the norm now. It’s not universal, there are plenty of holdouts and AI vegans, but as a software developer the expectation now is that you use some form of AI in your daily workflow.

Yes, you can point at the staggering environmnental impact and all the other ethical reasons why this is a terrible state of affairs. The industry, on average, does not care. That’s reality.

The reality is also that most teams and individuals are still grappling with the implications, and trying to find strategies and coping mechanisms, to make sure that the individuals, the teams, and the projects they work on, actually benefit from this technology they are now expected to use. And with the expectations of magical productivity that come with it. In that regard it’s still a wild west out there.

With acceptance comes commitment, finding a way to engage with reality that aligns with your values. What that means for you, is up to you. I can only encourage you to lean into that metacognitive awareness. To ask "are we being smart about this", and to reflect deeply upon the answer.

Bookmarks

The book links above use bookpile.org, an open source project that takes a book identifier (an ISBN), and shows you local bookshops and libraries. It’s a small independent project by Bodo Tasche, who I know from my Berlin Ruby days. It’s perfect for when you want to link to a book, without sending people to Amazon, Goodreads (owned by Amazon), or some other marketplace.

During my vacation I did take the time to watch a few conference talk recordings, one that stands out is this keynote from NDC Copenhagen, by Richard Campbell: After the AI Hype: What’s Real, and What’s Next. It does a fairly good job at providing a zero-bullshit recap and overview of where we are at.

With talks like these which paint a picture of the world, it’s always interesting to think about what’s being left out of the picture. Aside from a brief mention of DeepSeek, the Chinese AI industry gets ignored. It’s a common omission, both in North America and in Europe, but it’s no small oversight. A lot is happening in China, and it’s going to have a big impact globally in years to come. One of my favourite podcasts about China is Sinica. The host, Kaiser Kuo, besides a musician, intellectual, and China watcher, is also a technologist who was part of China’s early internet boom in the nillies, and frequently has guests that help him explore China’s technological developments. Two recent episodes can give you a taste of what’s brewing in China: Samm Sacks and Paul Triolo reporting from the 2026 World AI Conference, in Shanghai, and Angela Zhang and Alex Yang, talking about how China acts as a "platform state", with many parallels to technology platform companies.

Support

This is a reader supported publication. If you have the means please consider supporting me financially.

You can also help by forwarding this email to a friend, or sharing the link on your socials. Thank you!