Current AI State for me

5.3 won for me

OpenAI Plus sub, give me just enough usage for all my projects.

Also ampcode.com amazing free tier powered by ads. (10$/day of usage frontier models)

pi.dev - simple, minimal, slick. Very good primitives, just saying openclaw using it via sdk author: https://x.com/badlogicgames

Right now my stack is hybrid by default:

• one reliable paid model for day-to-day coding
• one free-tier heavy option for experiments
• one lightweight tool for quick ideation and drafts

What changed in early 2026 for me is not just model quality, but workflow quality. The biggest gain is less context switching and faster iteration loops.

How i use AI daily:

• scaffold boring parts fast
• review my own code with a second pass
• generate alternatives when blocked
• convert rough notes into structured tasks

Still, there are limits:

• long sessions can drift from original goal
• benchmark claims are noisy in real life
• tool UX still matters as much as raw model power

My rule now: use AI as force multiplier, not decision maker. I keep architecture and product calls manual, and use models for speed + coverage.

What i am watching next:

• pricing pressure on premium tiers
• better local/on-device options
• stronger agent tooling around repo-wide context