Ai-Assisted Porting Outcomes And Cost/Time Metrics

Issue 86 Edition 2026-03-27 4 min read

General

Sources: 1 • Confidence: Medium • Updated: 2026-04-13 03:54

Key takeaways

The write-up claims a first working Go version was built in 7 hours with approximately $400 in token spend.
The write-up claims the key enabling factor for the vibe-porting effort was JSONata's existing test suite.
The team used a one-week shadow deployment running old and new versions in parallel to confirm the new implementation matched the old behavior.
The author states that the “saved $500K/year” framing is somewhat hyperbolic.
The write-up frames the project as rewriting JSONata with AI in a day and saving $500K per year.

The write-up claims a first working Go version was built in 7 hours with approximately $400 in token spend.
The write-up frames the project as rewriting JSONata with AI in a day and saving $500K per year.
The project is presented as a vibe-porting case study that produced a custom Go implementation of the JSONata JSON expression language.

The write-up claims the key enabling factor for the vibe-porting effort was JSONata's existing test suite.

The team used a one-week shadow deployment running old and new versions in parallel to confirm the new implementation matched the old behavior.

What is the underlying cost model behind the $500K/year savings framing (what costs existed before, what changed after, and over what time horizon)?
How much additional engineering time and token spend was required after the initial 7-hour working version to reach production readiness and maintainability?
What is the scope and quality of the referenced JSONata test suite (coverage, types of tests, oracle quality), and how directly did it translate to the Go implementation’s confidence?
What criteria and metrics were used during the one-week shadow deployment to determine behavior matching (exact output equivalence, tolerances, performance, error handling)?
What compatibility surface was required for this JSONata reimplementation (Node-RED integration expectations, jq-like semantics, edge-case compatibility) and where did it diverge, if anywhere?

AI assisted porting plus a strong test suite may compress time to first working rewrite, lowering rewrite barriers for mature libraries.
Shadow deployment as a validation pattern may be becoming a standard de risk step for swapping behavior sensitive components.
ROI narratives around AI rewrites may be overstated without cost breakdowns, implying investor attention should shift from headline savings to verifiable engineering metrics.

Disclosed breakdown of the claimed annual savings, including baseline costs, post change costs, and time horizon, aligning with the author’s caveat about hyperbole.
Reported total effort from first working version to production readiness, including additional engineering time, token spend, and ongoing maintenance burden.
Quantified verification quality, such as test suite coverage and shadow deployment equivalence metrics, demonstrating reliable behavior matching beyond anecdote.

Evidence that production readiness required substantial additional time or rework, making the 7 hour result not representative of total cost.
Shadow deployment or post deployment issues show meaningful behavior divergence or edge case incompatibilities, undermining the validation approach.
Savings claims cannot be substantiated with a transparent cost model, reinforcing that ROI framing is not decision grade.