FIELD NOTE / X
The demo is the cheapest part.
The short film, the complete written thought, and the evidence behind it.
The X edition will be linked here after its public post is verified.
The demo is the cheapest part.
Video caption
The demo is the cheapest part. The owner leaves their actual job to run repair loops. Default to maintained tools for commodity work. #EricFieldNotes
Full written post / accessible read
Giving a capable employee a coding agent can turn a small need into a working prototype by Friday. That feels like saved money. But the spend starts when users depend on it and every failure needs diagnosis, re-prompting, regression tests and another release.
In an illustrative internal app, each new edge case sends its creator back through the agent: reproduce, explain, patch, retest, redeploy, then ask what else changed. No one necessarily hand-recodes the app. The expensive resource is the person's focused attention.
Before approving a build, name the service owner, users, data sensitivity, failure cost and replacement plan. Compare one maintained SaaS subscription with a measured month of internal support, including review and incident time. Keep the agent option only if it clears that full-cost bar.
Build the piece that expresses a genuine company-specific advantage. Buy routine scheduling, CRM plumbing or forms when maintained products solve them well. Do this because ten half-owned internal apps can consume more capacity than one useful engineer ever had to spare.
#EricFieldNotes
X thread draft
Video belongs on the first post; the rest gives the full argument. Check the native composer before sending.
The demo is the cheapest part. The owner leaves their actual job to run repair loops. Default to maintained tools for commodity work. #EricFieldNotes
Giving a capable employee a coding agent can turn a small need into a working prototype by Friday. That feels like saved money. But the spend starts when users depend on it and every failure needs diagnosis, re-prompting, regression tests and another release.
In an illustrative internal app, each new edge case sends its creator back through the agent: reproduce, explain, patch, retest, redeploy, then ask what else changed. No one necessarily hand-recodes the app. The expensive resource is the person's focused attention.
Before approving a build, name the service owner, users, data sensitivity, failure cost and replacement plan. Compare one maintained SaaS subscription with a measured month of internal support, including review and incident time. Keep the agent option only if it clears that
full-cost bar.
Build the piece that expresses a genuine company-specific advantage. Buy routine scheduling, CRM plumbing or forms when maintained products solve them well. Do this because ten half-owned internal apps can consume more capacity than one useful engineer ever had to spare.
Evidence and boundary
On-screen label: ILLUSTRATIVE COST MODEL. Illustrative cases are not measured incidents. Research papers and vendor documents support the stated mechanism only within their studied or documented scope.