Since opus 5.5 the scale of unreasonable-ness at @nsavecom has shifted gears. A few things that have helped:
- Lowering the barrier of prompting insanely. Most Slack messages go through an always-on listener and get scored on whether our AI can help or not pretty accurately at less than $5/day.
- Building an internal harness (of harnesses) which works relentlessly and really well. AND helps us in routing to cheapest option(s) available. We have used this harness since Jan
- Focusing on team dynamics and how we work such that everyone is authoring PRs (that are not slop grenades for the most part)
- Cheap testing environment with full agent control, quick uptime, prod parity, test as much as you like
This has a dramatic felt effect on our services from increased reliability, less customer support tickets, faster shipping features, less than a day on partner integrations, etc. All of this would have been impossible at this speed and efficiency 2 months ago. Opus 5.5 (and GLM 5.3) will go down as more pivotal than Fable/Mythos/Sol.
Next is completely overhauling PR reviews.
First of many. 🚀
Happy to announce that my first app is officially live on the Apple App Store and Google Play Store.
One more project in my portfolio💪🔥
Unemployment will humble you.
You will be an experienced software engineer and still end up taking a customer service job just because you don't want to be called lazy or someone who has nothing doing.
Scary, man.