iiinova — where your digital self keeps evolving

Simon Wardley and I are writing a book in the open about it called "Rewilding software engineering". You can read the existing chapters at: moldabledevelopment.com

060

However, getting accurate and representative answers quickly replaces beliefs with facts, and paves the way for real engineering to flourish. In doing so, we might discover that the gains are substantial, and that our own abilities are worth retaining and enhancing.

100

It's a simple idea that has been already shown to work in TDD, and that can be expanded to any questions about a system including the non-functional aspects of it. This is not to say that GenAI is not useful. It certainly is.

100

But there is a way to augment our understanding by enhancing how we can get accurate, representative and explainable answers for questions about our systems. This can be achieved through contextual deterministic tools that are specifically built to provide answers for each question.

100

Given these facts, it is rather natural to look at the promise of AI as the only way out. However, this way leads to surrendering our ability to understand deeply.

130

It is rarely a subject of conversation, and, even when it is, the reading process itself is seldom challenged. It was writing that captured the attention while reading was silently accepted as costly and inevitable. Indeed, developers spent most of their time reading code since a long time.

First, this is not a new problem; it has existed for decades. The good news is that it has never been seriously optimized. When was the last time you heard developers discuss *how* they read code (that is, not the code they read, but how they do the reading)?

Kirill Osenkov @kirillosenkov.com · 12/07/2026

"Software engineering has changed radically. It is now mainly about reading code. But reading code does not scale. We need to delegate to AI." I am glad you now recognize code reading as a problem. However, the broader problem statement is not quite correct and neither is the conclusion.

software engineering has not changed radically.

it has always been mainly about understanding code.

but reading code does not scale.

we need to augment our understanding through contextual tools.

183

Reposted by Tudor Girba

This made me dig up how I did drag and drop for C# statements in my old structured editor from 2006 - you could only drag them where the grammar allowed it github.com/KirillOsenko...

1302

Reposted by Tudor Girba

andrew blinn @disconcision.com · 11/07/2026

all will become fidgetable

1630437

Reposted by Tudor Girba

Beth Andres-Beck @beth4ma.bsky.social · 21/06/2026

To make AI work for all of us we should: 1. Stop the Slop Factories 2. Make It Safe 3. Fix The Taxes I’ll get it done: bethfordemocracy.com

12910

I have the privilege of working with extraordinary colleagues and amazing clients. That's it. It's a real privilege.

030

Moldable Development is a way of programming through contextual micro tools.

What should this investment look like? Start by reading the Rewilding Software Engineering chapters that @swardley.bsky.social and I wrote in the open, where we argue that software engineering can be optimized for decision making. moldabledevelopment.com 6/

moldabledevelopment.com

Moldable Development

030

While such an investment does not have a moat (nor should it), enhancing human abilities is very much in line with European values. U.S. could do the same, but given their current trajectory, they likely would consider it uninteresting and weak, which is a good premise for any asymmetric move. 5/

And therein lies the opportunity: invest in human intelligence. This investment implies new skills, new tools, and later, new processes. I should say that the investment is not at odds with an AI investment. It complements it. But it is many times less expensive and has only upsides. 4/

120

The hype behind Mythos is based on the argument that we have reached the peak of human ability and that AI has already surpassed it. Yet, when it comes to software engineering, we are far from even knowing what the peak of human ability is. 3/

Hopefully, it will lead to more investment in local and open-source models. This should have been the strategy a long time ago, but better now than later. Still, there is another less obvious lesson here. 2/

120

Tudor Girba @tudorgirba.com · 11/06/2026

The U.S. act to selectively allow Fable access based on nationality should make it finally obvious that an overly deep dependency on foreign actors is a threat to digital sovereignty. Hopefully, this can act as a wake-up call for Europe 1/

1100

I have seen complains that fable kills "slopus" as a term. I do not understand that position: fable does a lot of the work by itself as it literally means "fictional story ... that features animals, legendary creatures, plants, inanimate objects, or forces of nature that are anthropomorphized"

010

Tudor Girba @tudorgirba.com · 09/06/2026

The first thing we need to do when working with an LLM agent is to define the context. The more specific the better. But how specific can it be? How about a single object? (Here I am exploring the actions of a website visitor and then ask it to add a view for that action to see the relevant page)

000

Reposted by Tudor Girba

andrew blinn @disconcision.com · 08/06/2026

Participants will work through a combined tutorial on value probes and functional programming in Hazel, and then complete some writing and debugging tasks; 2 hours total. Due to time constraints basic functional programming experience is required. Interest form here: docs.google.com/forms/d/e/1F...

docs.google.com

Live Probes in Hazel: Study interest form

Thank you for your interest in our study. We are evaluating Live Probes, a new feature in the Hazel programming environment that displays runtime values inline in source code. As part of our tutorial ...

1202

Reposted by Tudor Girba

andrew blinn @disconcision.com · 08/06/2026

We’re doing a user study to find out how adding always-on live values alongside code effects writing and debugging experiences for functional programming. See reply for sign-up details

317338

Reposted by Tudor Girba

Konrad Hinsen @khinsen.net · 04/05/2026

This is a nice illustration of what I called "redressing the balance" in an essay for Onward! 2024: dl.acm.org/doi/10.1145/... Very different vocabulary, same message: we have been focusing too much on code as features and thereby neglected code as a knowledge representation.

dl.acm.org

Redressing the Balance: A Yin-Yang Perspective on Information Technology | Proceedings of the 2024 ACM SIGPLAN International Symposium on New Ideas, New Paradigms, and Reflections on Programming and S...

021

People talk these days about the importance of controlling the context. Indeed, I see it the same way. However, the context does not have to start from an external Markdown file. It can start from the knowledge already encoded in the system. 9/

000

To guide the AI interaction, I script it contextually. I explore what the AI does and see the results live. The agent works with my object and changes the code based on that context. And it is treated as just another object viewed in the system in the same way as anything else. 8/

To do that, I have to discover how the API models these connections and extend my model so I can navigate them. That does involve creating code, but my conversation starts from the very object I am inspecting, from deep inside my system. 7/

100

But how does GenAI fit into the view of systems as knowledge work? Take a look here. I am exploring an API; the GitHub API in this case. I arrived at a contributor entity, and now I want to see what this contributor is linked to. 6/

Most GenAI focuses on producing features. We pass in code, and it generates more code. Because understanding the inner workings is secondary, it is tempting to delegate that understanding to the machine. A console is enough, even preferable. 5/

Elegance is often a guide, and every time a simplification happens, a leap becomes possible. Any system can be seen in terms of features, and any system can be viewed as encoded knowledge. It is up to us. The choice influences what we praise. 4/

100

The view of a system as encoded knowledge is quieter. Evolving knowledge is less linear. The work is a continuous search for new ways to describe the space. When a new way is discovered, the system is reshaped to match it. The goal is to express more with less. 3/

Features are to be produced. Productivity gets measured in throughput. It can be seen in tickets that materialize in code instructions that manipulate data. The faster the better. The more the better. Force is often preferred. 2/

100