Why systems for humans, not people?
Main article
Why Systems for Humans, Not People?
This is the piece. The argument: software engineering and information systems are collapsing into one new practice. Behavioural engineering. The naming is unsettled because the scale is unprecedented. The practice itself is already running at planetary scale.
It walks through Coca-Cola as the original behavioural engineering project, the Ubuntu adage reread through a data-points lens, why the indie hacker question in Zimbabwe should be what does the system need around this want? rather than what do people want?, and finally the agency question. When the systems are this good at predicting us, what's left of the thing humanism puts at the centre of everything?
I don't answer it. The next few issues are me trying to.
The technical companion
How I built my Today and Tomorrow daily briefing skills
If the headline piece asks the question, this one is me trying to answer it in code.
I built two Claude skills that pull together my calendar, email, Slack, and iMessages into a single daily rundown. The whole architecture is built around a simple constraint: the system serves the human, not the other way around. No engagement optimisation. No notification baiting. No retention logic. Just useful information, fetched once, presented cleanly, then it gets out of the way.
It's a small thing. But it's a working example of what systems for humans looks like when an actual person ships one.
What I'm reading this month
A few pieces landed in May that all converge on the same anthropological question from completely different angles.
Magnifica Humanitas — Pope Leo XIV's first encyclical
Released on 25 May. The Pope's first encyclical, on "safeguarding the human person in the time of artificial intelligence." His framing: "The challenge we currently face is not technological, but anthropological."
That's almost exactly the argument my piece makes from completely different premises. I'm a humanist atheist. He's the Pope. We landed on the same question from opposite philosophical traditions. The convergence is the signal.
His framing of two cities is worth the read on its own. The Tower of Babel: a single language, a single technology, a single direction, built without reference to God, that ends in dispersion. The walls of Jerusalem under Nehemiah: rebuilt piece by piece, with each family assigned a section, through shared responsibility. He's asking: which one are we building?
I'll be unpacking this in Issue 4.
The AI-SDLC Framework — Ona
ona.com/stories/ai-sdlc-framework
Lucas Valtl at Ona has put out the sharpest framework I've seen for understanding where software engineering is actually going. Three stages.
Stage 1: In the loop. Developer writes the code, AI assists. This is the Copilot stage. Most organisations are here.
Stage 2: On the loop. Developer stops writing code and starts orchestrating multiple agents in parallel. The skill shifts from can I write this? to can I decompose, delegate, and evaluate?
Stage 3: Autonomous loops. Agents coordinate with each other. The human sets direction and validates outcomes.
The framework's real value isn't the maturity ladder. It's the bottleneck shift. Each stage moves the constraint somewhere new. Stage 1 makes code creation fast, so review becomes the bottleneck. Stage 2 makes execution fast, so planning and decomposition upstream and CI/CD downstream become the bottleneck. Stage 3 makes everything fast, so the bottleneck becomes human judgement at the top.
Two Google talks on the future of the craft
I watched How Software Development Has Changed and The Future of the Craft this month. Google's view: human-machine collaboration is the centre of software development now. Not human replacement. Not pure automation. The collaboration is the discipline.
The framing matters because it pushes back against both extremes that dominate the discourse — the AI is coming for everyone's jobs doom narrative and the AI changes nothing dismissal. Both are wrong. The thing that's changing is the shape of software work, not whether software work exists.
Also caught What's New In Flutter. Less philosophical, more practical. Notable for anyone shipping cross-platform apps from the Global South.
Micro1 — the new frontier of AI job interviews
Tried this out of curiosity. It's an AI-powered job interview platform. Everything automated. Conversational, engaging, surprisingly good at follow-up questions. This is what behavioural engineering looks like applied to hiring: a system that profiles candidates through interaction itself.
For developers in places like Zimbabwe, this is mostly good news. The barrier to being seen by global employers just dropped. The interview no longer needs a recruiter in San Francisco. But it's also exactly the systems-as-evaluators dynamic my essay was wrestling with. The candidate is now a data stream being read by an algorithm. The algorithm decides whether you proceed.
Worth trying. Worth thinking about.
Coming in Issue 4
The agency question I left open in Issue 3 is the thread for next month. I want to engage Magnifica Humanitas properly. There's a piece on how publishing platforms enclose — how Hashnode moving its GraphQL API behind a paywall is the same pattern as every other platform Medium-ifying itself over time.
— Mxo
