Design

Designing for Speed: How GigaToken Influences UX

When infrastructure gets dramatically faster, the experience you are allowed to design changes with it.

By Kiran · 5 min read · 26-07-2026

Speed is not a technical detail that lives in the backend. It is a design material, as real as color or type, and it shapes how an interface feels to use. When a piece of infrastructure gets dramatically faster, the experience you are allowed to design changes with it. GigaToken, a Rust tokenizer that encodes text at gigabytes per second, is a sharp example of exactly that. Designing for speed means understanding how those gains ripple all the way up to the user, and designing to spend them well.

What GigaToken actually changed

GigaToken is a byte pair encoding tokenizer built in Rust that runs language model tokenization at up to 24.53 gigabytes per second, reported as hundreds of times faster than the tokenizers most AI teams use today. Tokenization is the unglamorous first step in any AI product, the moment where text gets chopped into the units a model reads. It is easy to ignore because it is invisible. But invisible work that happens on every single request adds up, and speeding it up by orders of magnitude quietly loosens a constraint that touches the whole experience.

That is the interesting part for designers. When a bottleneck collapses, the question is no longer can we make this fast enough. It becomes what should we do with the time we just got back.

Response time thresholds and how each range feels to a user. Icons: Clock, Stopwatch, Hourglass
Users do not measure milliseconds, but they feel every one. Faster infrastructure pushes more of the pipeline into the instant range. Icons: Clock, Stopwatch, Hourglass.

Speed is felt as trust

Users do not measure latency in milliseconds, but they feel it instantly. A response that arrives the moment they ask reads as competence. A delay, even a short one, plants a seed of doubt. Research on interface responsiveness has been consistent for decades: under roughly 100 milliseconds a system feels instantaneous, up to a second it feels connected but noticeable, and beyond that attention starts to wander and trust erodes.

Infrastructure like GigaToken pushes more of the pipeline under those thresholds. As a designer, your job is to make sure the interface actually cashes in that speed rather than squandering it on unnecessary spinners, staged animations, or artificial delays added in the name of polish.

Three ways faster infrastructure changes the UX you can design

1. Real time becomes the default, not the premium tier

When preprocessing is nearly free, features that once felt too expensive to run on every keystroke become ordinary. Live suggestions as someone types, instant reformatting, results that update while the query is still being written. These interactions feel magical precisely because there is no wait. Faster infrastructure moves them from a special feature to the baseline expectation, and your designs should assume that baseline.

2. You can show more, sooner

Speed lets you replace loading states with content. Instead of a spinner while the system thinks, you can stream partial results, show a live preview, or render an optimistic version immediately and refine it. Every spinner you can delete is a small gift of confidence to the user. The fastest interface is the one that never makes the person wonder whether anything is happening.

3. Iteration loops get tighter

When each cycle is cheap, users try more things. They experiment, undo, and refine without the friction of waiting punishing every attempt. Designers can lean into this by making actions reversible and low commitment, encouraging play instead of careful, hesitant clicking. A fast system invites exploration; a slow one teaches caution.

Do not spend all your speed in one place

Here is the trap. A faster backend tempts teams to fill the newfound headroom with heavier features until the experience is slow again. Speed you win at the infrastructure layer is easy to lose at the interface layer through bloated pages, oversized assets, and unnecessary work on the main thread.

Treat performance as a budget you protect, not a windfall you immediately spend. Some concrete habits help. Keep your critical path lean so the first meaningful moment arrives fast. Optimize the assets that ship on every screen, since images and icons are common culprits; lightweight, consistent visuals like the soft 3D icons in the Thridy gallery render cleanly without dragging the page down. And measure real experiences, not just benchmarks, because users live on the ninety fifth percentile, not the average.

Design the perception, not just the number

Even with fast infrastructure, perceived speed is its own craft. A few principles carry most of the weight.

  • Respond immediately, even before you are done. Acknowledge an action the instant it happens with a state change, so the user knows the system heard them.
  • Show progress honestly. If something genuinely takes time, a determinate indicator beats an endless spinner, because a visible finish line feels shorter than an unknown one.
  • Prioritize what the eye lands on. Load and render the content the user is looking at first; defer the rest.
  • Never add fake delay. Artificial pauses meant to feel deliberate almost always read as slow instead. Trust the speed you have.

The bigger lesson for AI products

GigaToken is one component in a long chain, but it points at a pattern that will define the next wave of AI interfaces. As each layer of the stack gets faster, the winning products will not be the ones with the most features. They will be the ones that feel instant, where the technology disappears and the person stays in flow. Speed, used well, is invisible. The user never thinks about tokenization or latency. They just feel that the tool keeps up with their mind.

That is the real design opportunity in every speed gain. Not to do more, but to make what you already do feel effortless. If you are building AI powered experiences, treat every millisecond you win as attention returned to the user, and design like that attention is the most valuable thing you have, because it is. Browse the Thridy category index when you need fast, consistent visuals that keep your interface light.

Free 3D icons for this

Every icon on Thridy is free to download and use. A few that fit this piece:

Browse all 2,600+ free 3D icons or explore them by category.

Key takeaways

  • Treat speed as a design material that shapes how an interface feels.
  • When a bottleneck collapses, ask what to do with the time you got back, not just how to go faster.
  • Faster infrastructure makes real time the default and lets you delete spinners.
  • Protect performance as a budget; do not spend every gain on heavier features.
  • Perceived speed is craft: respond instantly, show honest progress, and never fake delay.

Questions people ask

What is GigaToken?

GigaToken is a byte pair encoding tokenizer written in Rust that runs language model tokenization at up to 24.53 gigabytes per second, reported as hundreds of times faster than commonly used tokenizers. It speeds up the invisible first step of AI text processing.

How does faster infrastructure change UX?

It moves real time interactions from premium features to the default, lets you replace loading states with streamed or optimistic content, and tightens iteration loops so users experiment freely instead of hesitating through waits.

Can you lose the speed you gain?

Easily. A faster backend tempts teams to fill the headroom with heavier features until the experience is slow again. Treat performance as a budget you protect, keep the critical path lean, and optimize assets that ship on every screen.

Does perceived speed still matter with fast infrastructure?

Yes. Acknowledge actions instantly, show honest progress, load what the eye lands on first, and never add fake delay. Perceived speed is its own craft that sits on top of real performance.

Sources

This article was produced with the help of generative AI, drawing on the sources above, and reviewed by a human before publishing. All 3D icons shown are original artwork from Thridy. If you spot anything that needs a correction, let us know.

Enjoyed this article?

Subscribe to the Thridy Journal. One great story at a time, no noise.

Keep reading