App Spotlight — bixso.academy: an AI meter that charges what the model actually costs
[hook] Meet bixso.academy again — this time not for its two AI agents, but for how it prices them. Most platforms charge one flat number per AI answer, no matter which model actually did the work.
[model] bixso.academy runs several AI subagents behind the scenes: some do light feedback, two do heavier reasoning (career advice, work grading). Charging every one of them the same Energy price means the platform either overcharges the easy questions or undercharges the expensive ones — there's no honest middle.
[analysis] Model first, then the rule, then the automation: only the two subagents that need real reasoning call the heavier model; the rest run on a lighter one that costs a fraction as much for most tasks. Energy price now follows which model actually ran — a heavier call costs more Energy than a light one — instead of one flat rate for every AI answer, deployed live.
[agentic] The accounting underneath had quietly split into two systems that stopped agreeing: an admin could top up a user's Energy and it wouldn't change whether their AI worked; a user could run out of AI and their visible wallet balance wouldn't move. Fixed by wiring both to the same ledger, so the balance a user sees is the balance that actually gates them — plus the 100⚡ signup grant, written months earlier but never wired to fire, now reaches new users automatically.
[mirror] None of this is dramatic — it's plumbing. But for anything that charges for metered AI, "the number you show equals the number that governs you" is the whole trust proposition. Worth asking of your own product: does your meter and your bill actually read the same number?
[door] Live at bixso.academy — book a free first consult if your product needs AI pricing that's honest by construction, not a flat guess dressed up as a feature.
bixso.academy · Created by BIXSO
Open in bixso.net →