Microtokens, Cheap Compute & A Quieter, Faster Functory

The problem

Functory’s pricing model was starting to hit a wall.

As cheaper compute tiers came online, the old “whole token” billing made everything feel either too coarse or too hacky.

Charging 0.005 tokens per second is fine on paper, but in reality it meant rounding, confusion, and a feeling that the meter didn’t quite match what the machine was doing.

At the same time, the new design was beautiful but not entirely honest: the footer carried heavy, continuous animations that looked nice in short demos but made the experience feel sluggish over longer sessions — especially on less powerful machines.

The product wanted to feel sharp and lightweight. The foundations weren’t fully there yet.

What shipped

Two small but meaningful shifts:

1. Microtokens for precise, cheap compute

• Added support for microtokens so billing no longer has to pretend everything costs a whole token.
• Functory can now bill very low-cost tiers with sub-second precision while still keeping a clean internal accounting model.
• This unlocks a future where “small, fast functions” on modest hardware feel natural — and still properly tracked.

2. A lighter, calmer interface

• Tracked down a performance issue to a heavily animated footer that was always running, even when it wasn’t adding much value.
• Simplified the animations and reduced the workload on the browser, especially on long pages.
• The result: less noise, less hidden cost for the user’s CPU, and a UI that feels more respectful of attention and hardware.

None of this radically changes the screenshots. But it changes how the product feels.

Why it matters

Microtokens are not a marketing feature. They’re a quiet statement:

“We’re taking precision seriously, even when the numbers are small.”

If Functory wants to host a world of small, composable utilities — scripts, tools, agents — it needs a billing layer that doesn’t flinch when something costs fractions of a cent.

Microtokens make it possible to ship cheap, honest compute without hand-waving.

The UI work follows the same spirit.

Nice animations are easy. A fast, quiet UI that stays fast after 100 runs, 50 tabs, or an old laptop… that’s harder.

Fixing the footer is a small step toward a product that chooses clarity and speed over visual ego.

How we built it

• Extended the billing logic to support micro-level units while keeping the mental model of “tokens” for humans.
• Ensured existing data can be interpreted consistently, so old runs and new runs still make sense side by side.
• Verified that pricing for new cheap tiers flows correctly from model config → compute tiers → tokens → microtokens → invoices.

On the performance side:

• Profiled the new layout until the culprit became obvious: a footer that was always doing work in the background.
• Reduced animation complexity and frequency, keeping a bit of personality but cutting unnecessary strain.
• Re-tested on multiple browsers and machines to make sure the change actually makes the page feel lighter, not just look different in dev tools.

What’s next

• Keep tightening the billing model as more compute tiers appear, always pushing for fairness and transparency rather than cleverness.
• Add better internal tools to inspect run costs at microtoken level, so it’s easy to see where time and money go.
• Continue hunting down small performance leaks in the UI: anything that wastes the user’s CPU or attention is worth revisiting.

Still a tiny product, still early — but these quiet updates are the kind that let Functory scale without drifting away from its promise:

simple, honest infrastructure for people who just want their code to run, and to be paid fairly when it does.