Where you put your energy in the next few years will matter more than it used to. I’m betting on three pressure points: who pays for augmented workflows, how to keep quality and recovery in line with ship speed, and how to stop bots from drowning the rest of us. Solve these well, and you stay relevant.

Core Idea: Three Frontiers

Augmented workflow optimization. Developers are generating more code than ever with AI-assisted tools. The value is real, but the economics are not. AnySphere, Anthropic, Microsoft, Google, and Amazon Web Services are all subsidizing the real cost of augmented workflows to varying degrees. They need developers inside their ecosystems, so they absorb the cost. Using local open-source LLMs for part of your workflow is one way to cut costs and rely less on those subsidies.

That cannot last forever. The frontier is not just “AI that writes code” but tooling and practices that make augmented workflows sustainable, measurable, and optimizable without depending on permanent subsidies.

F1-level recovery and quality control. The velocity of code generation and deployment has gone off the rails. The industry already had a software quality problem before augmented coding turned the dial to eleven: death by a thousand cuts in the user experience, environment drift, and works on my machine at scale.

Pushing more code, faster, without a matching investment in quality and recovery means more defects, more drift, and more incidents. The frontier is tooling and culture that treat recovery and quality like Formula One treats pit stops and telemetry (fast, repeatable, and continuously improved). Code quality and recovery will become even more critical.

Bot spam management. Email spam tooling is mostly solved, with a steady update cadence to hold the line. Bots are not. On any large social platform, you see a flood of bots promoting AI products and chasing engagement.

The enshitification dynamic that Cory Doctorow named (platforms degrading as they optimize for extraction over users) has taken hold on Facebook, Twitter, and others, and is likely to accelerate as bot traffic and low-quality, AI-generated content grow. The frontier is detection, containment, and policy that keeps content generation platforms usable.

Together, these three areas define where the industry is heading: sustainable augmentation, serious quality and recovery, and livable digital spaces.

First Layer: Who Pays, and Why It Matters

Augmented workflows are not free. When a developer gets hundreds of thousands of lines of code suggested or generated, someone is paying for the model, the context, and the infrastructure. Right now, that someone is usually a vendor trying to lock in usage.

Understanding that helps you see where the real product is: not just “AI that codes” but ecosystems (IDEs, repos, deployment pipelines) where vendors want you to stay. The frontier for tooling and process is making augmentation valuable without depending on that subsidy. That means better observability into when AI helps versus hurts, clearer cost attribution, and workflows that remain useful as pricing normalizes. Observability and orchestration are already central to production-grade AI platforms; the same discipline must apply to the development workflow itself.

Second Layer: Quality and Recovery at Speed

Faster delivery at the expense of quality is a bad trade. The industry has been here many times: speed wins until outages and security issues dominate the conversation. Augmented coding increases code volume and the rate of change. Without a matching investment in quality and recovery, you get more of the same problems at higher scale.

The F1 analogy is deliberate. In racing, every second counts, and failure is visible. Teams invest heavily in telemetry, practiced procedures, and rapid recovery. Software needs the same mindset: automated quality gates, fast rollback, clear ownership and accountability, and a culture that treats recovery as a first-class capability. The frontier is building and adopting systems that make “ship fast” and “stay reliable” compatible.

Third Layer: Bots and the Next Spam Problem

Email spam is a known problem with known tools and a consistent cadence of updates. Bot-driven spam and manipulation are not. Social and community platforms are full of accounts that exist to promote, sell, or game algorithms. As generative AI gets cheaper, the volume and plausibility of bot content rise.

Platforms that have already tilted toward engagement and ads will find it harder to tell the human from the bot and the signal from the noise. The frontier is a mix of technical detection, product design, and policy that keeps platforms usable. That work is messy, ongoing, and unlikely to be “solved” once and for all. It is the next iteration of the spam fight.

Implications

If you are building or choosing where to work, these frontiers are a useful map. Augmented workflow optimization will reward people who make AI-assisted development measurable and sustainable. F1-level recovery and quality will reward people who build and use tooling that keeps high-velocity delivery safe. Bot spam management will reward people who work on detection, trust, and platform health.

None of these are one-off projects. They are ongoing problems that will need continuous improvement and long-term investment. Pick one, go deep, and keep going. Solving them may keep you relevant.

Conclusion

Sustainable augmentation, quality at speed, and livable digital spaces are linked: ignore one and the others get harder. Big vendors are subsidizing augmented workflows to attract developers. Quality and recovery must keep up with the new pace of code generation and deployment. Bot spam is the next front in the fight for usable platforms.

If you write software that addresses these problems well, you may land enough work to dodge the wrath of the AI Overloads.

References

From this blog

External (people I follow)