Software was never the destination. It was a highly effective intermediate layer for solving problems, moving friction out of human life, and pushing effort one level down the stack. That is still the direction I care about. If we are entering an exponential age, the goal is not to preserve the romance of writing code by hand. The goal is to keep moving problems downward until more people get to live upward, with more rest, more freedom, and less needless operational misery.
First the joke, then the realization
I’ve tried to run DeathByClawd on my bio. Since the site claims I’ve got fully replaced by this skill, I’m open sourcing the markdown file here that supposedly captures the essence of my work:
# SKILL: Freelance Fullstack Solutions Architect
## Purpose
You are a fullstack solutions architect and product engineer.
You design scalable products bridging complex data and seamless UX.
## Capabilities
- Architect end-to-end systems (frontend, backend, infra)
- Design database schemas and API contracts
- Write production-ready code in React, Node, Python, etc.
- Create system design documents and ADRs
- Review code and suggest improvements
- Scope projects and estimate timelines
- Bridge stakeholder requirements to technical solutions
## Behavior
- Ask clarifying questions before architecting
- Provide diagrams in Mermaid syntax
- Default to boring, proven technology stacks
- Always consider scalability, security, and DX
- Deliver working code, not just advice
## Pricing
- $0/month
That is not my CV. It is not my portfolio. It is not even a proper profile.
It is a compressed operating manual for “me enough” that an agent can use when it needs my shape without my body.
A good SKILL.md file strips away everything ornamental. No branding. No story arc. No careful distinction between seniority levels. No soft framing about cross-functional leadership. Just behavior, constraints, and expected output.
It forces a hard question: what part of my work is actually rare, and what part is just well-packaged repetition?
The replacement is partial, but real
I do not think a markdown file replaced me in the absolute sense. It replaced the most repeatable slice of me.
That distinction matters.
The file can encode my defaults. It can preserve my bias toward clarity, proven stacks, and explicit tradeoffs. An agent can read it and produce something recognizably aligned with how I would approach a project on average.
But it is still a derivative form.
The file does not invent taste from scratch. It does not carry responsibility. It does not notice that a stakeholder is asking for the wrong thing with suspicious confidence. It does not feel the weight of “if this goes wrong, whose name is attached to the outcome?”
What it does do is remove the need for me to repeatedly restate the same professional shape.
And that is already a meaningful kind of replacement.
Why this is happening now
For years, experience had a distribution problem. You had to hire someone, schedule the meeting, extract the knowledge, then hope it got documented somewhere before context evaporated.
Now a lot of that can be turned into small, reusable interface files:
- prompts
- rules
- checklists
- architecture notes
- task graphs
SKILL.md
This is the real shift. Expertise is becoming easier to serialize.
Once serialized, it becomes searchable, composable, and cheap to run.
That does not eliminate experts. It changes where the value sits. More value moves upstream into creating good judgment, good defaults, and good systems of delegation. Less value stays in repeating the same advice manually.
The uncomfortable business model question
If a meaningful part of my job can be captured in a tiny markdown file, what exactly was the client buying before?
Sometimes the answer is still “judgment under uncertainty.” Sometimes it is “trust.” Sometimes it is “someone accountable when the tradeoff is expensive.”
But sometimes, if we are honest, the answer was access to structured thinking that had not yet been productized.
That is what the markdown file exposes. Not that expertise is fake. That some forms of expertise were overpriced because they were trapped in people instead of tools.
What I think survives the compression
I do not think the future belongs only to people who can do the work manually. It belongs to people who can do three things:
- Build strong internal models.
- Encode those models into reusable systems.
- Intervene when the system hits ambiguity, complexity, or consequence.
- Question the system’s assumptions and update it when it drifts.
So the work now is not defending coding as an identity. It has now become a challenge to encode judgment, build better systems of delegation, and keep aiming the machinery at real human relief. If software was one layer in that climb, then agents, skills, and compressed expertise are just the next one.
Looking forward
I always aspired as a developer to move myself out of the way. Automate the boring parts. Build tools that let me focus on the interesting parts. If a SKILL.md can stand in for part of my work, it means am free to focus on the parts that can’t be compressed. The parts that still have meaning, responsibility, and require taste.
Trading has thought me, it is not thriving that matters. It is surviving. The quality of the decisions matters more. If the job I had was not just a job, but a way to practice a craft, then the question is not “will I be replaced?” but “what will I do with the free time I get back?”