In 2022, I wrote that DevOps had become waste.

The response was predictable: “DevOps can’t be waste - we need automation!”

They missed the point.

DevOps principles were right. But when every team rebuilds the same CI/CD pipeline, writes the same Terraform modules, and solves the same problems in isolation, that’s not DevOps. That’s local optimization.

[Read: DevOps is the New Waste in 2023 — briancarpio.com]

Platform Engineering emerged as the fix. Centralize infrastructure so teams focus on product, not YAML.

It worked. For infrastructure.

But now we have a new wave of waste. And it’s bigger than DevOps ever was.

The New Waste: AI Local Optimization

Teams everywhere are now:

• •Writing custom prompts for AI code generation
• •Rebuilding context injection into LLMs for every microservice
• •Manually reviewing and correcting LLM output
• •Using LLMs like magic vending machines — no feedback loops, no lessons retained

I see this every week as a consultant.

One team at Company A spends two sprints wiring up AI to understand their codebase — breaking it into chunks, injecting docs into prompts, trying to make Claude or GPT give better answers.

Meanwhile, Company B is doing the exact same thing. Different team, same goal, same half-baked playbook. Neither knows the other exists. Neither is capturing what works. Both are reinventing how to give AI context from scratch.

And most orgs?

They’re still stuck trying to pick Copilot vs. CodeWhisperer vs. Windsurf — with zero plan to measure impact or build repeatable systems.

This is the 2025 version of “every team writes their own Jenkins pipeline.”

It’s AI local optimization at scale. Thousands of teams figuring out how to use GPT or Claude or Bedrock… in isolation.

No shared context. No reinforced outcomes. No alignment to business goals.

OutcomeOps is the answer.

OutcomeOps: The Next Evolution

OutcomeOps is the cultural evolution that DevOps was always pointing toward.

It’s not about shipping faster. It’s about aligning every system — human and machine — around measurable outcomes.

Where DevOps unified Dev and Ops, OutcomeOps unifies engineering + AI + product + architecture under a single operating philosophy:

• •Outcome over output.
• •Feedback over fire-and-forget.
• •Augmentation over automation.

It says:

• •Don’t write prompts — write repeatable thinking systems
• •Don’t guess what works — log, measure, iterate
• •Don’t just ship AI-generated code — co-engineer with it
• •Don’t let every team reinvent the same tool — codify the pattern once, share everywhere

That’s what Context Engineering is all about — structuring knowledge, memory, and constraints so LLMs think with your team, not at them.

[Read: The OutcomeOps Way: Stop Prompting, Start Co-Engineering]

History Repeating

Most companies will miss it again — just like they did DevOps.

They’ll buy AI tools. They’ll mandate LLM usage. They’ll track “AI adoption” instead of AI impact.

The winners will be the ones who treat AI as an engineering system — not a chatbot.

That’s OutcomeOps.

Want to see OutcomeOps in practice?

I’ve open-sourced the system I use: outcome-ops-ai-assist

It ingests your ADRs and code-maps, generates code that follows your standards, and validates output before merge.

Learn more at outcomeops.ai.