How we build
ASEWAVE
Build with AI speed. Ship with human-grade evidence.
Augmented Solo Engineering With Adversarial Verification + Evidence
The method in 30 seconds
Think of it as a building site with independent inspection: AI agents build fast, a second agent checks the work without trusting the first, and nothing is signed off until an experienced master has seen the evidence – tests, screenshots and a change log.
Developed for the AI era.
Wi24rd develops its solutions with ASEWAVE: Augmented Solo Engineering With Adversarial Verification + Evidence. It is a disciplined, human-owned, AI-assisted development method built around independent verification, evidence trails and replayable review.
AI agents can write code fast. They can also write confident fiction fast. ASEWAVE is designed for an environment where AI may increase pace but must never become an unverified source of truth.
One developer owns scope, priorities and every destructive decision. AI agents assist with specification, implementation and verification – but their claims are checked against the actual codebase by another agent that does not trust the first.
The result is an audit trail by construction.
Not just “the feature works” – but what changed, why it changed, how it was tested, and what evidence supports the claim: commit hashes, test output, screenshots, file paths, diffs and reproducible checks.
The principle is the same as in a penetration-test exam: a finding only counts when the proof can be checked.
The model
Three phases, a human gate after each.
Each unit of work moves through three phases. The phase boundaries are process gates, not optional checkpoints.
Investigate
Read the codebase, verify assumptions and produce an investigation report. No code is written yet.
Implement
Execute the approved spec in rounds, capturing artefacts and test output as you go.
Verify
An independent check of claims, followed by a walkthrough the human reviews before the phase closes.
Four roles that don’t trust each other
Human Gatekeeper – the master
Controls scope, priorities and releases, and approves every gate. Makes every destructive decision.
Reviewer / Prompt Author
Writes the phase prompts and challenges unsupported claims. Does not write production code.
Executor Agent
Implements approved work in its own session, captures artefacts and produces evidence.
Evidence Verifier
Checks that claims match primary evidence: file existence, unique hashes, test output.
The letters
Augmented
AI increases output. Human ownership keeps direction, scope and judgment grounded.
Solo Engineering
One developer owns the product, priorities, architecture and every destructive decision.
Adversarial Verification
Independent verification checks claims against code, commits and test evidence before changes are accepted.
Evidence
Commit hashes, screenshots, test output and walkthroughs – not narrative claims alone.
What ASEWAVE is not
Honest about the limits.
- Not autonomous AI development – there is a human decision at every step.
- Not “AI-assisted coding” in the loose sense – phase artefacts, tests and walkthroughs are mandatory.
- Not faster than a senior team – but more than one person could otherwise deliver with documentation.
- Not a fit for pure prototyping or single-file scripts.
Public and private
The public method and the whitepaper are documented at asewave.org. The operational engine (prompt contracts, verifier logic and similar) is private and not released.
Wi24rd FTP, tingsvad.dk and XDRagon Monitor are all built with ASEWAVE. Wi24rd FTP went from first commit to v1.0.0 in 6½ hours – every commit is public.
Read the whole method.
ASEWAVE is documented openly at asewave.org – roles, phases and red flags.