Dev log.

Are your reps setting the time, and wrapping up the call properly? Despite the last post, I've been thinking about what "MVP" actually means for Calleoke. Not minimum viable — minimum sellable. What's the smallest thing someone would pay for? My first instinct: build the teleprompter. Real-time coaching, karaoke-ball tracking, suggestions that evolve as confidence grows. The full vision. But that's a lot of complex moving parts. And complex means slow. Slow means no customers. No customers means no feedback. No feedback means building blind. So I asked a different question: what's the core value, stripped to its simplest form?

The Insight

The teleprompter tells you what to say. But before that matters, there's something simpler: Did you hit the key objectives?

Every sales call has moments that matter — did you confirm the duration of today's meeting, did you agree on next steps in the call, did you book the next meeting and send the invite and confirm receipt. Miss one of these and the deal stalls. Not because you said the wrong thing — because you forgot or didn't have time to say the right thing.

The MVP

So the MVP isn't the teleprompter. It's the objectives tracker. Listen to the call. Track what matters. Alert when something critical is about to be missed. "You're 20 minutes in and haven't confirmed next steps." Simple. Concrete. Valuable today.

Why This First

• It's the foundation anyway — you need to understand what happened before you can suggest what to say next.
• Faster to build — no complex real-time UI choreography. Just listen, track, nudge.
• Easier to sell — "Never forget to book the next meeting" is a problem people know they have.

The teleprompter is where I'm going. The objectives tracker is how I get the first check.

Next

Find the right pilot customer. Someone with real calls, real objectives, real stakes. Their process becomes the blueprint. We build it together.

If that sounds like you — let's talk.

I've been heads-down on what makes real-time coaching actually feel right. Not just "AI that helps with sales calls" — that's too vague. I needed principles that guide every design decision. Here's what I landed on:

1. Partner, Not Supervisor

Most tools feel like a manager watching over your shoulder. Calleoke should feel like a brilliant colleague in your ear.

When you go off-script — and you will, you're in the conversation, the AI isn't — Calleoke doesn't correct you. It adapts. You took things somewhere unexpected? Here's help for wherever you ended up.No "you should have said X" energy. Just collaboration.The same coaching works for everyone. Experienced reps glance and riff. Newer reps follow more closely. One product, used differently.

2. Thinking Out Loud

Most AI: wait for input → process → show answer. You see "thinking..." then suddenly a result. Feels slow and dumb, even when it's fast.

Calleoke shows its thinking as it develops. Customer starts talking, you see: "Hmm, sounds like a concern..." Customer continues: "Looks like pricing..." Customer finishes: "Price objection — here's how to show ROI."Confidence grows visibly. Not a loading spinner. A window into Calleoke's mind working alongside yours.

3. Tracking, Not Blocking

When you're delivering lines, the teleprompter should feel like a karaoke ball — alive and tracking with you, not frozen and waiting.Lines light up as you speak. They fade when done. Next line promotes smoothly. If new suggestions need to appear, they queue up and arrive at natural pauses, not mid-sentence.The whole thing breathes.

The Test

Every UI decision gets checked against these three. Does it feel collaborative? Does it show thinking? Does it track without blocking?Yes? → Ship it. No? → Rethink. Hit me up if you want to see the latest build.

We’ve been cooking in the editing room and are putting together some cinematic promos!

Still raw cuts but final edits and more “real-world” type stuff coming soon.

Working on the front-end and the various types of interactions, see video below.This video is a “pre-recorded” demo intended to show the user flow and different customer situations and how this would display to the user. this front-end up to the back-end, should end up in a usable product, albeit for “constrained” situations.model refinement and tuning: experimenting with new models and timings to improve speed and feel. Promo videos: also working on cinematic promo videos, will be cool.Currently in a good rhytm where I’m able to get some work done on this everyday, besides all my other obligations. Make sure to check back in, or reach out below.

This release is intended ONLY to showcase the concept behind Calleoke (and my vibecoding skills), it is still far from actually being useful - there’s a plan on how to get there tho.Known issues:- Transcripts showing up as both customer and rep - this is just a feedback issue, today you can simply resolve it by using headphones. Complete fix planned.- Speed and timing: yes it’s slow and timings are off. these are addressable by simply tuning various timings and logics. additionally currently using pretty basic models with many intresting models with appeal (including apples latest).- UI: yes there’s more refinements and complexity in the UI/UX, planned, see concepts in link below.- Quality of Lines: of course, but we’re not even there yet. There’ll be multiple levels of intelligence running at the same time and plenty of refinements.

The first working prototype is taking shape.

Real-time transcription ✓

Speaker identification ✓

Dynamic prompting... in progress.

Everything changed. AI capabilities exploded. Similar solutions started appearing in adjacent spaces—validating that the world was ready. New tools like vibe-coding meant a non-technical founder could actually build sophisticated prototypes. So I stopped waiting for a technical co-founder and started building it myself.

Armed with nothing but frustration and a slide deck, I pitched this vision to technical founders. The conversations were promising but ultimately fizzled out. The idea was there. The path to build it wasn't.