Malcolm
Warm and curious. He turns what you say into how you’d want to say it.
Everyday English with Malcolm
Listen on SpotifyEvidence
We only show what we can prove.
Every claim has a date and a state: works in the app, works in part, too early to call, or not published. If anything here doesn’t match the app, tell us and we’ll fix it.
Four states
- Works today
- Check it in the app yourself.
- In part
- It works, but not everywhere — or it’s not on screen yet.
- Too early to call
- We don’t have the data, so we claim nothing.
- Not published
- We don’t do this, and we don’t pretend to.
What you can check in the app today
What you’ll see on your own phone.
Screens from a demo account with a sample conversation history. The screens are real; each caption says only what’s on it.
“What I remember about you”: see what your AI coach has kept and where it came from — correct or delete any item. 
Your learning plan: reorder or remove any step.
Claim by claim
The evidence table.
Nobody outside FizzUp has checked any of this yet — the “Checked by” column says so. Where there’s no sample, we give no number.
| What we claim | State | Date | Method | n | Checked by | How to check it yourself |
|---|---|---|---|---|---|---|
| Four AI coaches run a conversation in different ways | Works today | 26 Sep 2026 | The same situation with each of the four AI coaches | — | Internally; nobody outside | Pick an AI coach when you start. |
| We show what your AI coach knows about you, and you can delete it | Works today | 26 Sep 2026 | Your AI coach’s memory screen: delete single items or everything | — | Internally: code and app review | Profile → “What I remember about you” (latest version of the app) |
| You can have the same conversation in text, without the microphone | Works today | 26 Sep 2026 | A live conversation with a text box; Role Play started by typing, no microphone | — | Internally | In a conversation: the transcript tab has a box to type in. In Role Play: “Start by typing” (latest version) |
| A phrase comes back on its review day | Works today | 26 Sep 2026 | A review schedule based on the SM-2 algorithm | — | Internally: code review | Calendar → Agenda |
| A correction that fits the moment: a better version, a nudge or — for now — nothing | In part | 26 Sep 2026 | A rule for choosing the kind of correction. In text conversations and the tools — in code; in voice conversations — an instruction to the model | No figure until we have a protocol | Internally, results not published | One kind — the better version — you can try in the widget at the bottom of this page |
| Progress counted when you use it on your own | In part | 26 Sep 2026 | A phrase comes back at least a day later, in a new conversation, before your AI coach says it; phrases that were on screen or in the scenario are discarded | Hand-checked sample — planned; we’ll give the size once it’s done | Nobody yet | Not visible yet: the “Used again” screen is being built |
| The CEFR level in the app | In part | 26 Sep 2026 | An estimate from your conversations; not validated | — | Nobody outside | Profile → Progress: a ring with the estimated level. Treat it as a rough guide |
| Fluency gains in real situations | Too early to call | 26 Sep 2026 | Comparable speaking tasks over time, blind-rated | — | — | The first group hasn’t started yet |
| Regular practice | Too early to call | 26 Sep 2026 | Active weeks | — | — | The first group hasn’t started yet |
| The Founding Companies pilot | Too early to call | 26 Sep 2026 | Criteria agreed with the company before the start | — | — | Not started yet. I’ll show you the criteria when we talk about a pilot |
| An effectiveness study | Too early to call | 26 Sep 2026 | A pilot study; we’ll publish its plan before it starts | — | — | In preparation |
| Points for pronunciation | Not published | 26 Sep 2026 | Pronunciation Lab is a preview with no score | — | — | Tools → Pronunciation |
Prepared, not switched on yet
This is how we count progress: only when you say it yourself.
At least a day later, in a new conversation, before your AI coach prompts you. Repeating it right after you’ve heard it doesn’t count — and neither do minutes or taps. The measurement is prepared, but it isn’t switched on yet. The “Used again” screen that will show it to you is being built.
Hear the AI coaches.
Each of them hosts a podcast in the same voice they use with you in the app.
Each AI coach hosts a podcast in the same voice.
Theo
Direct and quick. He won’t let you hide behind simple sentences.
Interview Ready
Listen on Spotify
Emma
The precision and tone that get you taken seriously.
Diplomatic English
Listen on SpotifyRules
Rules we stick to.
- Openly AI. Your AI coach doesn’t pretend to be human — that’s how we designed it.
- A company sees group results, never one person’s mistakes. Groups that are too small aren’t shown.
- Recordings only with your consent, and only briefly; transcripts are kept for a limited time.
- Your data doesn’t train AI without your separate consent.
- We don’t promise results. Minutes are limited — we don’t promise endless conversation.
We don’t list retention periods in days here — the exact figures are in the Privacy Policy.
Want to try this with your own team?
Let’s talk about a pilot ›Try it on your own sentence
Say it better.
Type a sentence you’d say in English at work. Malcolm shows you a better version — and why. No account needed, up to 10 sentences a day.
Or start from one of these
Openly AI. No pretending to be human. Your sentence isn’t stored.
The conversation ends. The learning doesn’t.
Start with the conversation you most want to get right.
Download FizzUp
The app is free: no subscriptions, no in-app purchases, no ads.