How to use Claude to create Calcapp templates
A Calcapp co-founder used Claude Code to create 20 templates in a weekend, supplying sample JSON and custom formula syntax, then correcting Claude's understanding of LET support.
news.ycombinator.com
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
Use [sample Calcapp JSON] to generate [template type] for [domain]. Follow these formula extensions: [supported syntax]. Use LET where helpful, including [separator rules], and ask about unsupported functions before choosing a more verbose formula.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source poster / publisher
- davidpolberger
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- The author reports creating 20 new Calcapp templates in a weekend, with most generated in one attempt and one formula refined after clarification of LET support.
- Human involvement
- The author inspected a generated formula and clarified supported syntax; the post does not describe a formal test or release review.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- The author prepared a sample JSON file for Claude Code to examine.
- The author explained Calcapp's extensions to Excel-style functions.
- Claude generated templates for real estate, financial planning, and retail.
A Calcapp co-founder used Claude Code over a weekend to create 20 new templates for the company's formula-driven app builder. The starting reference was a sample JSON file the author prepared. After examining it, Claude generated apps that the author describes as competent and intricate, with most produced in one attempt.
The author also explained Calcapp's extensions to Excel-style functions. These included lambdas for FILTER and named sort-type enums for XMATCH. They report that Claude incorporated those extensions immediately. The templates covered several domains, including real estate, financial planning, and retail; the author credits Claude's domain knowledge with helping produce them.
One exchange led to a specific revision. Claude generated a verbose formula and remarked that LET support would make it cleaner. The author clarified that Calcapp already supported LET and allowed := between names and values. Claude then rewrote the formula using that extended syntax.
The reported result is a batch of templates built in a couple of days. Integrating Claude Code into Calcapp's own designer remained an idea the author was holding off on because of cost concerns.
Who it's for: App-builder teams that can provide representative structured examples and explain their platform's formula extensions.
Prompt
Suggested prompt (written by Agents In Use, not the author's)
Source: Hacker News · davidpolberger
Frequently asked questions
What reference did the author provide?
The author prepared a sample JSON file and briefly explained Calcapp extensions such as FILTER lambdas and named sort-type enums for XMATCH.
How did Claude improve the verbose formula?
After the author explained Calcapp's LET support and := separator extension, Claude rewrote the formula using that syntax.
More details
Remaining steps
- The author clarified LET support and Claude rewrote a verbose formula.
About this account
Claude Code generated most apps in one attempt from sample JSON, while the author supplied platform details and corrected a formula-syntax assumption.
Platform: Hacker News. Attribution: https://news.ycombinator.com/item?id=45725421. Prompt: Suggested (written by Agents In Use, not the author's). Sourced by the Agents In Use pipeline.
Credit: davidpolberger on Hacker News
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · Calcapp · JSON
Calcapp · Templates · Formulas
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.