Heads up : overage for github copilot is (no longer) FREE

Tools and Libraries
1

Just letting people know that currently copilot paid versions (which is cheap as chips) have FREE overage. I’m clocking $10 a day in overage fees (for thousands of requests) that I dont have to pay while it’s in preview (not really preview IMHO, but anyway.

In other words you can use it until you hit daily rate limits (which is per LLM, not for whole service) like a mad man.

@Zackster I’m sure you and the core team are all over this, but use it while it’s free to generate out for instance jdbc extensions for lucee with admin pages by pointint it at some existing repos and saying, now create me one for duckdb or ____

2

ask it anything, put vscode extension in ‘agent’ mode and then script out your database schemas, and ask it questions about the quality of your code, where you can improve performance, or if there is better ways to build functions, sql stored procs, functions etc…

I’ve milked this baby dry for two weeks, and it’s INSANE how much domain knowledge on frameworks, coding patterns, code review and alike we’ve done.

We’ve probably achieved about two years of coding in two weeks, if we hired a 30 year vet of cfml/sql/js/css and built out a tonne of stuff, so we’ll be contributing a whole range of lucee extensions, and cfml components in the coming months to the community to add functionalities to apps.

you can even do security reviews of your code, setup, everything.

we dont expose data to the AI, but sometimes we generate out scripts dumps of our data to inform decisions in our apps as we build them with ‘help’ from the AI.

Dont forget also to create (or get AI to create) instruction files (github instruction files) one for each framework or tool that you’re using staring with best practices, and add in-house coding guidelines, then anything else you need the ai to know about your app in an app based instruction file and you’re off.

BE CAREFUL NOT TO EXPOSE SENSITIVE DATA, otherwise code wise it’s all a comodity really. You shouldn’t have usernames and passwords in code in 2025 anyway, but just thought I’d point this out clearly before people expose their code (and accidently sensitive info) to an AI.

Sorry, just been super excited by the results from this ai, we use Claude Sonner 3.7 (they have 4.0 in preview, but 3.7 is great), you can also use 3.7 thinking in ‘ask’ mode (on left hand side of chat) to think deeper, but that mode can not look at your workspace, only answer questions so you’ll have to copy and paste, but sometimes great for conceptual answers.

Anyway, get a free code review! and keep asking more and more questions of your code.

1 Like
3

Here’s a list of models I’ve found useful and and overview:

o4-mini: The Lean Strategist

  • Design Philosophy: Prioritizes clarity and modularity. It’s great at breaking down requirements into clean, reusable components.
  • Strengths:
    • Fast iteration cycles — ideal for prototyping secure flows or UI logic.
    • Handles encryption workflows and modular Ext JS setups with minimal overhead.
  • Limitations: May not anticipate edge cases or long-term scalability unless prompted explicitly.
  • Best For: Architecting lightweight, scalable systems where performance and cost-efficiency matter.

:brain: Claude Sonnet 4: The Systems Architect

  • Design Philosophy: Thinks in layers and abstractions. It’s like having a senior dev who maps out the entire ecosystem before writing a line of code.
  • Strengths:
    • Excels at multi-service orchestration, API design, and domain-driven architecture.
    • Great at reasoning through security models, especially for browser-based encryption.
  • Limitations: Can overengineer — sometimes adds complexity that’s elegant but not strictly necessary.
  • Best For: Architecting enterprise-grade platforms, especially with complex data flows or policy constraints.

:brain: ChatGPT 4.1: The Creative Engineer

  • Design Philosophy: Prioritizes developer experience and usability. It’s intuitive and fast, but sometimes skips architectural rigor.
  • Strengths:
    • Fantastic for UI/UX flows, theme generation, and user-centric design.
    • Can rapidly scaffold apps with modern frameworks like Lucee or Ext JS.
  • Limitations: May miss deeper architectural concerns like data integrity or long-term maintainability.
  • Best For: Architecting front-end-heavy apps, MVPs, or creative tools like WYSIWYG editors.

:wrench: Summary Table

Model Architecture Style Best Use Case Risk Profile
o4-mini Modular & efficient Secure, scalable microservices May under-specify
Claude Sonnet 4 Layered & robust Enterprise systems, encryption flows May over-specify
ChatGPT 4.1 Intuitive & creative UX-first apps, MVPs, design systems May oversimplify
1 Like
4

Gemini for vscode has a huge free tier too

5

Gemini 2.5 pro in copilot, (not direct) didn’t seem to code as well with cfml for me? Got any tips?

HOT TIP

Btw, set a $5 budget on your github copilot account and you can get MORE than the limit of requests, remember OVERAGE IS FREE…well to a point. so when you hit the limit, you dont, as it charges yo… see pics, asking helpdesk for clarification on this.

image
image1217×622 26.3 KB

image
image1232×871 28.4 KB

Yeah I maxed it out in 10 days :wink:

6

As of today no longer free