iiinova — where your digital self keeps evolving

I talk SWE @italkswe.bsky.social · 01/10/2026

Agents not being able to consistently follow a simple red, green, refactor loop should be evidence enough that “long running autonomous agents” is a bad idea.

000

I talk SWE @italkswe.bsky.social · 30/09/2026

Of all the XP practices forgotten/overlooked be mainstream SWE culture, empathy is the one I miss the most

000

I talk SWE @italkswe.bsky.social · 27/09/2026

The best software design teacher I ever had was TDD. Instead of passively sitting around waiting for experience to accumulate by it self, TDD expedites learning. Future software engineers should be trained in manual (canon) TDD before moving on to agent based workflows.

000

I talk SWE @italkswe.bsky.social · 18/09/2026

There is no increase in productivity without an equally increase in comprehensibility.

010

I talk SWE @italkswe.bsky.social · 17/09/2026

Hard pill to swallow: The average engineer has never heard about XP, TDD, TBD, refactoring, simple design, dependency inversion, ports and adapters or continuous delivery 😔

000

I talk SWE @italkswe.bsky.social · 12/09/2026

There are some noticeable parallels between agentic guidance and leadership theory. Is all about creating an environment which doesn’t just enable the right outcomes but make them inevitable

000

I talk SWE @italkswe.bsky.social · 12/09/2026

@birgitta410.bsky.social a fun kind of sensor we’re trying out in my team is visualizing architecture by hexagonal “roles” by analyzing the codebase (deterministically) through introspection. This results in an interactive graph of use cases, ports and collaborators.

100

I talk SWE @italkswe.bsky.social · 12/09/2026

My current workflow is to work at the Hexagonal Architecture level of abstraction first and then define acceptance tests in small coherent batches.

000

I talk SWE @italkswe.bsky.social · 12/09/2026

For years I (too) would be pedantic about TDD being “once small test” at a time. But “gearing” was always part of the equation and the agents do allow bigger bolder steps.

000

I talk SWE @italkswe.bsky.social · 12/09/2026

“Brevity is the soul of wit” I’m looking at you Fable and GPT-6

001

I talk SWE @italkswe.bsky.social · 03/09/2026

TDD is still extremely crucial when working with agents. Tests or small batches of tests are perfect at strong prompting

000

I talk SWE @italkswe.bsky.social · 03/09/2026

Working effectively with agents is all about decomposing work in to comprehensible, verifiable, working slices which can be committed straight to main.

000

I talk SWE @italkswe.bsky.social · 29/08/2026

I pi as my agent harness primarily so that I can add to APPEN_SYSTEM.md ‘’’ - us impersonal language ‘’’ I do not want to in a parasocial relationship with my work tools

000

I talk SWE @italkswe.bsky.social · 29/08/2026

The real bottleneck is still (and will always remain) *what* problems to solve

000

I talk SWE @italkswe.bsky.social · 29/08/2026

Software quality (e.i ability to change safely) is absolutely not dead. If anything it is more important than ever. If “code is cheap” then customers should expect more for less. This creates quality pressure.

000

I talk SWE @italkswe.bsky.social · 26/08/2026

Are people seriously going all stacked diffs rather than just do TBD?

000

I talk SWE @italkswe.bsky.social · 26/08/2026

Working in small batches of coherent acceptance tests gives a nice balance between speed and confidence

000

I talk SWE @italkswe.bsky.social · 24/08/2026

There is a lot more nuance to agentic development than the zeitgeist allows for

000

Reposted by I talk SWE

joesteel.bsky.social @joesteel.bsky.social · 24/08/2026

Tell me again how your carefully curated specs can mitigate this? Good fundamentals, good tests (ideally with mutation testing) are absolutely essential. Even then, you're not fully covered. The genie can be helpful, but always be careful what you ask for and verify what you get. It won't be liable

002

I talk SWE @italkswe.bsky.social · 23/08/2026

At its heart software development is all about creative problem solving. Agent are a true force multiplier but they are no replacement for the raw humanity at the center of the creative process

021

I talk SWE @italkswe.bsky.social · 23/08/2026

Long running autonomous swarms of agents is a recipe for disaster

000

I talk SWE @italkswe.bsky.social · 22/08/2026

Agents are really bad at storing skills for agents

000

I talk SWE @italkswe.bsky.social · 22/08/2026

If a skill defines new agent behavior it should list each distinct piece of guidance instead of Hemingwaying in between the lines

000

I talk SWE @italkswe.bsky.social · 22/08/2026

Many SKILLs read like prose which I don’t think optimal. It makes it hard to change in a structured way.

000

I talk SWE @italkswe.bsky.social · 22/08/2026

In my team we are seeing a tremendous boost to effectiveness with the help of agents. Not by swarming or automation but by empowering people to work better. Improve collaboration with more informed conversations, greet higher quality feedback with many more rapid prototypes.

000

I talk SWE @italkswe.bsky.social · 20/08/2026

I’ve really come to appreciate how crucial TDD was in teaching me to externalizer my thinking

000

I talk SWE @italkswe.bsky.social · 19/08/2026

People are (still) the moat

000

I talk SWE @italkswe.bsky.social · 17/08/2026

We’ve been improving our agentic guidance by replacing verbose descriptions with “common conceptual shorthands”

000

I talk SWE @italkswe.bsky.social · 16/08/2026

Well established patterns such as hexagonal architecture provide a wonderful off-the-shelf ubiquitous language with the agents

000

I talk SWE @italkswe.bsky.social · 16/08/2026

I’m bullish on Software Engineering fundamentals

000

I talk SWE @italkswe.bsky.social · 16/08/2026

Agents can be a real force *multiplier* not *addition*

000

I talk SWE @italkswe.bsky.social · 16/08/2026

The most revolutionizing aspect of XP was not the testing or even the focus on feedback. It putting the humans in the center of the development workflow.

000

I talk SWE @italkswe.bsky.social · 15/08/2026

The single strongest pattern in all of software engineering is teamwork.

000

I talk SWE @italkswe.bsky.social · 23/07/2026

Being surprised by what an agent produces is a failure mode

000

I talk SWE @italkswe.bsky.social · 19/07/2026

I’ve been updating some old (6 months) slopOps skills. From 3.5k lines of meandering nonsense to 224 lines of highly specific actionable guidance.

000

I talk SWE @italkswe.bsky.social · 16/07/2026

Our current agentic workflow defines work outside-in. First a *why and what* document targeting the PM and the engineers. Then a high level *how* document describing in hexagonal terms how this charge can be expressed in our codebase targeting engineers /1

100

I talk SWE @italkswe.bsky.social · 15/07/2026

I truly hate the anthropomorphisation of LLMs. I don’t want to be in a parasocial relationship with my work tools. It’s nothing but a cynical engagement trap and the people responsible should be ashamed of them selfs

010

I talk SWE @italkswe.bsky.social · 15/07/2026

How do you classify which problems can be solved by autonomously by an agent and which require human judgement? Most (if not all) problems has a mix of both. Even a “perfect” agent would still need feedback in order to correct wrong assumptions

000

I talk SWE @italkswe.bsky.social · 13/07/2026

Hexagonal architecture is not highlighted enough as a core agentic guide. It does a lot of heavy lifting wrt separation of concerns, responsibility spilts and exposing the right seams for testing. In my team we also incorporated some ideas from functional core/imperative shell.

100

I talk SWE @italkswe.bsky.social · 13/07/2026

“No Silver Bullet” is more relevant than ever. The agents are good at handling the accidental complexity but helpless with the essential complexity

000

I talk SWE @italkswe.bsky.social · 11/07/2026

The thing I love the most about software development is that joy and productivity go hand in hand

010

I talk SWE @italkswe.bsky.social · 11/07/2026

Much of what I hear and read about long running agent jobs sound like software development alchemy

000

I talk SWE @italkswe.bsky.social · 11/07/2026

Pro tip: it’s easier than ever to create a project specific linter

000

I talk SWE @italkswe.bsky.social · 11/07/2026

Programming is a sub-branch of mathematics but software engendering is better understood as a sub-branch of epidemiology

000

I talk SWE @italkswe.bsky.social · 11/07/2026

I might be late to the party but @birgitta410.bsky.social work on harness engineering martinfowler.com/articles/har... is amazing! It is really helping shape the conversations about agentic work in my team

martinfowler.com

Harness engineering for coding agent users

A mental model for building trust in coding agents through feedforward guides, feedback sensors, and iterative harness engineering.

031

I talk SWE @italkswe.bsky.social · 08/07/2026

“Make the next change easy” is the core recursive algorithm of highly effective software engineering. The hard part is learning the difference between easy and fastest.

000

I talk SWE @italkswe.bsky.social · 08/07/2026

AGENTS.md is important but also extremely sensitive. Less is more

110

I talk SWE @italkswe.bsky.social · 07/07/2026

A key principle in my team when working with agents is *just-in-time*. As we refine plans from high level all the way down to tests we make sure to never let the agent get ahead of its skis. Define work one abstraction level at a time top-down then implement one vertical slice inside-out

000

I talk SWE @italkswe.bsky.social · 07/07/2026

My experience with agentic development: - many more much smaller sessions - maintain a common language with the agent - established architectural pattern (hexagonal) - UBIQUITOUS_LANGUAGE.md - behavior oriented tests are “strong prompting” - the LLMs are still worse than most people think

010