Practice Makes Progress

A year ago, if you’d asked me what would most renew my love of building software, I’d probably have given you a sensible answer: a cleaner architecture, a calmer backlog, a client who values craft as much as outcomes.

I wouldn’t have said: pair-programming with a tireless, occasionally overconfident, astonishingly useful set of coding agents.

And yet — here we are.

Over the past twelve months at Wundr Space, we’ve been living inside a new kind of development rhythm: working with large language models and coding agents (Claude and Codex in particular) as active collaborators in our engineering workflow. Not as novelty. Not as gimmick. As colleagues of a sort — co-piloteers — learning our patterns, challenging our assumptions, accelerating our throughput, and (when we’re doing it right) raising our standards.

It has been fun. It has been testing. It has been the most excited we’ve felt about programming in a long time.

A new team shape

The shift isn’t just “we write code faster now”.

The deeper change is that we’ve started to reimagine DevSecOps itself: not as a pipeline that humans feed, but as an AI-augmented team system — one where the shape of collaboration changes.

When you add capable agents into the mix, you suddenly have:

• a companion to explore options quickly (without dragging the whole team into every rabbit hole)

• a relentless assistant for refactors, scaffolds, test generation, documentation, and edge-case spotting

• an always-available reviewer who can interrogate decisions, surface risks, and propose alternatives

• a translator between intent and implementation — especially when the problem is clear but the route is messy

But (and this matters) those benefits aren’t automatic. They’re earned. They require practice.

Optimising for the things that actually matter

We’ve spent a lot of time tuning our workflow to keep the promise without inheriting the chaos.

Because speed without quality is just debt arriving early.

And creativity without security is just risk dressed as momentum.

So we’ve treated this year as a deliberate practice ground: optimising how we build to protect the outcomes we care about most:

• Quality — readable systems, not just working code

• Speed — delivery as a habit, not a heroic event

• Security — woven in, not bolted on at the end

• Scalability — technical and operational, so the work can grow with the vision

Sometimes that optimisation has felt like play. Sometimes it has felt like friction. Often it’s been both at once: the joy of possibility paired with the discipline required to make it real.

“Practice makes progress” (not perfect)

Along the way we found ourselves repeating a phrase — partly as encouragement, partly as a boundary:

Practice makes progress.

Not “practice makes perfect”.

Progress is the point. Progress is measurable. Progress is something you can share, learn from, iterate on, and improve.

Perfect is… a mirage. Or worse: a reason to delay value, hide work, and mistake polish for impact.

This framing sits neatly alongside a principle we’ve long respected: done is better than perfect.

But we’re not using it as an excuse to be sloppy. Quite the opposite.

We mean: value emerges through creative thinking paired with disciplined engineering practices — delivered in increments, tested in reality, improved in the open.

Progress is the pathway where quality actually gets built.

The synergy effect: we’re learning together

There’s another layer here that’s hard to ignore: the tools are improving at pace.

As we learn how to prompt, structure tasks, constrain outputs, and verify results, our co-piloteers get better at meeting us where we are. And as the underlying models improve, our techniques gain compound returns.

It’s a feedback loop:

• we become better users, designers, and reviewers of AI-assisted work

• the agents become better partners for our context and style

• the system as a whole becomes more capable than either side alone

That synergy is what makes this moment feel unprecedented — not simply because productivity is up, but because possibility is up. You can explore more approaches, test more hypotheses, and turn more ideas into working artefacts than traditional workflows could comfortably sustain.

And that creates a particular kind of energy: the feeling that the craft is expanding.

A new lease of life (and a higher bar)

In a quiet way, this has given us a new lease of life as programmers.

Not because we’ve outsourced thinking — but because we’ve reclaimed time for the parts of engineering that require taste, judgement, and care.

Agents can help us move faster through the mechanical layers. That creates more room for:

• better questions

• clearer designs

• sharper threat modelling

• stronger review culture

• more consistent test discipline

• more thoughtful operational readiness

• more deliberate alignment with the humans we’re building for

If anything, working this way has raised the bar — because the bottleneck becomes less about typing and more about engineering maturity.

The agent doesn’t replace the craft. It exposes it.

Where we’re heading

We’re still early in this. And we’re trying to stay honest about what we don’t yet know.

But one thing is clear: the future of DevSecOps won’t just be “pipelines with AI sprinkled on top”.

It will be teams and systems designed from the ground up for AI-augmented delivery — with new rituals, new interfaces, new safeguards, and new ways of thinking about responsibility.

For us at Wundr Space, this year has been a practice ground for that future.

A year of iteration. A year of learning in public (and sometimes in private, with a slightly embarrassed laugh). A year of building muscle memory for a new mode of creation.

And if we’ve learnt anything worth sharing, it’s this:

Progress is not the opposite of quality.Progress is how quality is born.

So we keep practising.

Not to be perfect.

To make progress — the kind that compounds.

---

Share: LinkedIn · Email