Can Bots make estimates as well as they code?

BLUF (Bottom Line Up Front)

Now for something completely serious. As the WordPress Whisperer post indicates, Bots help write code. But can they make estimates? ChatGPT with improved memory, has promised to remember a scale and give me a tee shirt size when I start a project. Let’s get estimates and find out how they compare with actuals.

Add a comment if you want to know how we do, I’ll keep you updated


One of my projects is to build a graphical editor for the SEH  app .

Goal

The purpose of SE-Editor is to enable editing all SE Processes for a System in a directed graph so that design intent and systems thinking is clear, and application sturcture and UI do not get in the way.

Notes

GraphViz produces better directed graphs than any other free/low cost tools I have found, it generates configurable svg which scales well. The main issue has been adding context menu driven by what the user clicks on. Angular and React offer context menus but attempts to integrate with GraphViz got nowhere. With help from ChatpGPT I was able to get a plain HTML, svg and javascript implementation to work, and provide an accessible means of extending functionality.

Implementation

1. GraphViz is invoked by a server method SystemReportServer which gathers data from Innovator, creates nodes and edges converts them with Dot2Svg and returns the svg.

2. Client method show_se_process_report4 writes HTML including svg to an iframe in Innovator UI

3. The HTML incudes

    – panzoom.js which handles scaling

    – se_editor4_1.js which creates the context menus

    – se_editor.css which styles the menus

4. Preferred iframe is for a sidebar menu button on sys_system ItemType

It works.

I asked for an estimate, Bot said 2 to 4 prompts . I told it the actual was 40. Well  that’s a pretty human underestimate. This is  where the scale above came from.

It’s like Deep Thought all over again. “You never really clearly said what the question was”.

The quest is now a voyage of discovery, with a chronometer as well as a compass.

We want AI to be fairy dust, just sprinkle some around and your problems go away. It’s more like a spoonful of sugar it helps the medicine go down. It’s 1 per cent fairy dust and 99 percent perspiration. 40 prompts is perspiration. It’s like riding a bicycle, you still have to press the pedals, and steer to avoid potholes an traffic. The payoff is you get further with the same amount of pedaling.

Leave a comment

Your email address will not be published. Required fields are marked *