---
title: Dan Walsh, log
last_updated: 2026-11-13
canonical: https://www.danwalsh.co/log
text_version: https://www.danwalsh.co/log.md
feed: https://www.danwalsh.co/log.xml
entries: 13
---

# Dan Walsh, log

I got a lot of thoughts trapped in some private repos and my brain.

So here's a log of thoughts, things I found interesting, little challenges, etc. that are too small to warrant full blog posts.

Newest first. A date of the form YYYY-MM means only the month is on record. Links where one exists. "backfilled" marks an entry written after the fact from repo and chat records; entries without it were written when they happened, from 2026-11-10.

## November 2026 (2026-11)

### 2026-11-13 · musing

Been thinking about the future of SWE and agentic use in general (shocker, who working in SW hasn't?)

In my eyes: verification and governance.

My prediction on SWE directions: verification engineering becoming a massive discipline/the new standard for software engineering at large.

We're no longer hand crafting pieces of curated, artisanal code to try and solve problems.

Aside: people sometimes seem to forget solving problems was the point of the job in the end - writing code in and of itself was not the full job, but many people seem to have made it their personality or their measure of value/impact as an engineer, and are now struggling with their identity as a result, but that's a different thread.

I understand the tension here. Hand writing code gave many of us the level of scrutiny that allowed for an ongoing understanding and verification in terms of the code produced, but as the be all and end all of a given engineer's identity, it's now far too limiting.

We're now at the factory's production line, and our job is to assess which of the different experiments and approaches output by an agentic session best address a given problem, which is most extensible and/or maintainable for the future and its problems to address, and to provide a level of confidence and rigour to how we arrived at that decision.

How do we do that? With the necessary guardrails, assessments, and verification practices in place.

Zoom out, same problem appears at a bigger scale. Verification is governance at the scale of a PR; governance is verification at the scale of an organisation.

Governance.

Not the sexiest option by a mile, but it feels like the right route, because we're slowly (or quickly depending on your personal perceptions and p(doom)) offloading our cognitive capabilities to a subscription service.

Intelligence and aptitude on demand, where we need to pay a given value each month to be in with a fighting chance in the ongoing arms race of individual capability.

I'm convinced that, of all the AI safety risks we need to guard against, gradual disempowerment and over-reliance are far too easily overlooked in favour of misuse, for example. Misuse / CBRN applications are obviously terrifying, and real problems, but the insidiousness of disempowerment is the killer to me.

In X industry, Y organisation is so much further ahead than us! We need to use AI to catch up. We don't have time to spend dwelling on these decisions and problems - DecisionBot 9000, what's your suggestion here? Ok great, implement that.

Fast forward Z months/weeks/days later - I don't understand any of this. Where are we? Let's ask AI - it's best informed and capable to consume the information we have here, get its assessment and then we'll go with its recommendations.

At what point do we recognise the problem that's developed? How do we even begin to pull ourselves out of that hole? Do we even recognise we're boiling along with the other frogs?

We need governance in place to try and offset these situations at the earliest possible points. In practice: named human(s) owning each significant decision, with the ability to explain it without asking a model, and recording why a given recommendation was accepted, not just that it was.

I don't wanna be a dang frog.

Links: [permalink](https://www.danwalsh.co/log#2026-11-13-future-directions-of-travel)

## March 2026 (2026-03)

### 2026-03-04 · published · backfilled

In my first blog piece, I quoted a study from METR (linked) around the impact of AI on experienced OS dev productivity in early 2025.

I want to revisit that. The quoted study was already out of date at the time I was using it, and it felt wrong not to address that.

So this was created to give more space to the study findings themselves, rather than using them to support a point, and some thoughts on where that could lead the SWE role itself in the future.

Links: [blog](/blog) · [Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity](https://metr.org/blog/2025-07-10-early-2025-ai-experienced-os-dev-study/) · [permalink](https://www.danwalsh.co/log#2026-03-04-perception-gap-rebuilt-around-own-usage)

### 2026-03-03 · decision · backfilled

This was an interesting one!

What does AI creativity look like?

This was the result of giving Claude full creative control on a piece, only requesting it stay within the (then present) text-only constraints of the pieces so far, and to write up an analysis piece as per the established convention.

I find this piece quite soothing in some ways, and the analysis is one I like to revisit from time to time.

This also raised an interesting thought around attribution - is this technically _my_ piece when I was expressly and vehemently hands off for as much as possible of its creation?

I decided no in the end - I just provided the infrastructure for it.

Interesting thought experiments.

Links: [intora.net](https://intora.net) · [CODEC analysis](https://www.intora.net/txt/sig/001/analysis) · [permalink](https://www.danwalsh.co/log#2026-03-03-sig-and-attribution)

## February 2026 (2026-02)

### 2026-02-28 · idea · backfilled

How do you best represent yourself when you come from a non-traditional SWE background, and have worn a fair few hats over your career?

I spent some ideating around the idea of `π-on-platform`, informed by the classic T-shaped developer archetype.

My thinking was - broad fullstack & infra knowledge providing the horizontal, the first vertical depth provided by frontend expertise, with the second vertical being  developing and deepening AI & agentic knowledge area.

Was an interesting one to think through and reflect on, but ultimately felt a little too abstract to be fully viable.

Links: [permalink](https://www.danwalsh.co/log#2026-02-28-on-platform-later-called-stale)

### 2026-02-23 · finding · backfilled

Saw a lot of noise online around infrastructure being a secure bet for  future-proofing in the era of agentic software.

Dug in a little - to some extents, likely correct, but in terms of orchestration and context engineering (at that point), rather than infrastructure ops itself.

This was mostly very vocal sellers trying to pitch their Kubernetes courses lol.
Dang you grifters!

Links: [permalink](https://www.danwalsh.co/log#2026-02-23-infrastructure-thesis)

### 2026-02-19 · built · backfilled

The second INTORA piece.

Number stations have some weird hold over me. They occupy the same 'mysterious and neat' vibes that lighthouses (for some reason) also give me.

I like how the sound turned out on this one. good job Tone.js.

Speaking of tone - also happy with the overall tone of the piece too. Still holds up to me.

Links: [intora.net](https://intora.net) · [STATION analysis](https://www.intora.net/txt/int/002/analysis) · [permalink](https://www.danwalsh.co/log#2026-02-19-int-002-station-first-audio)

### 2026-02-15 · built · published · backfilled

Ahhhh, my not-frequently-updated-enough art pieces.

The first! DRIFT. The site as a whole has evolved significantly since this first piece, but this was my first foray into this style of generative art.

What can be created with the constraints of text only characters? Can we replicate some natural flow patterns through this?

Links: [intora.net](https://intora.net) · [DRIFT analysis](https://www.intora.net/txt/int/001/analysis) · [permalink](https://www.danwalsh.co/log#2026-02-15-intora-net)

### 2026-02-12 · built · backfilled

Follow on from the AI usage retro - deriving some skills based on my own usecases.

Blog review - does the structure of a piece hold up?

Plan handoffs - I _love_ ideating in an AI chat, kicking ideas around, testing edge cases and verifying, iterating and planning things out.

How do you pass that to Claude Code to implement?

With a handoff skill! Codify how the plans are structured and output, and codify how they're consumed and actioned, including pre flight checks to ensure everything required for a given plan is accessible before beginning.

Still very fond of that skill, still gets a lot of use ~9 months on.

Links: [permalink](https://www.danwalsh.co/log#2026-02-12-claude-code-skills)

### 2026-02-11 · finding · backfilled

My first meta retro on using AI.

Running a retro through Claude.ai and Claude Code to uncover frictions, positives, and opportunities after 10 months of interactions.

Derived a playbook of patterns based on some of the patterns I'd started developing, alongside some interesting insights on context infrastructure being more useful for me rather than prompt refinements.

Links: [post](/blog/what-claude-taught-me) · [permalink](https://www.danwalsh.co/log#2026-02-11-ai-interaction-retro)

### 2026-02-11 · decision · published · backfilled

Ahhhh, the strategy repo. The enabler and bane of my life.

How do I keep track of my ideas, interests, current and future projects, potential blog posts?!

Of course -  a version controlled repo!

Links: [post](/blog/building-a-living-strategy-repo) · [permalink](https://www.danwalsh.co/log#2026-02-11-strategy-repo)

### 2026-02-06 · decision · backfilled

The aim of Momentum was to be able to access it from any Claude surface; web, desktop, mobile, Claude Code.

Turns out, the bearer auth approach I was using failed out when trying to use Claude.ai connectors.

Fell back to also using OAuth PKCE to handle those cases.

Links: [commit edc8fe5](https://github.com/dang-w/momentum-mcp-server/commit/edc8fe5) · [commit 8b21103](https://github.com/dang-w/momentum-mcp-server/commit/8b21103) · [permalink](https://www.danwalsh.co/log#2026-02-06-momentum-auth)

### 2026-02-05 · published · backfilled

Blog created and first post published. Ahhh, the horrors of being perceived and putting thoughts out there. Good times.

Links: [blog](/blog) · [permalink](https://www.danwalsh.co/log#2026-02-05-first-post-one-stat-re-attributed)

### 2026-02-05 · built · backfilled

Momentum! My first real foray into exploring the MCP protocol.

Momentum was designed as an always-accessible data store - thought process on how it was built was:

1. Go - hadn't got using it in production in a good few years, so felt like a good attempt to revisit and refamiliarise - also led to some experiments on the best way to absorb language patterns, conventions, and structures, while using agentic coding. Interesting one.
2. MCP - hadn't actually had any experience with MCP beyond consuming others' servers - felt like a good opportunity to actually trial it myself and assess.
3. Something I may actually use day to day. And I did! For a while. Classic lol

Links: [momentum-mcp-server](https://github.com/dang-w/momentum-mcp-server) · [permalink](https://www.danwalsh.co/log#2026-02-05-momentum)
