Muhammed Kandeger

Founder and entrepreneur. I build apps with a team of AI agents and sell like a B2B pro. Building in public toward freedom from the 9 to 5.

[Muhammed Kandeger @Mkandeger2]

@ashen3256711 One cheap test: inject synthetic transits into noisy light curves and see how many the model catches after detrending. It shows quickly whether the preprocessing is eating real signals.

[Muhammed Kandeger @Mkandeger2]

@guptankit27 Dialect gaps in search is a sharp angle, most brands only test one spelling of what people type. Breaking that case into a short thread with the before and after would be very saveable.

[Muhammed Kandeger @Mkandeger2]

@SandrineZambou @X Clarity first makes sense. If a chatbot can't sum up what you do in one line from your homepage, it won't recommend you either. Concrete proof beats adjectives for people and models.

[Muhammed Kandeger @Mkandeger2]

@ConnorTalksAI Sign language translation is a great pick. Are you reading signs live from the webcam or from recorded clips? Fast hand movement is where I'd expect it to get hard.

[Muhammed Kandeger @Mkandeger2]

@guptankit27 LinkedIn makes sense for a B2B tool. On X, showing the work tends to beat pitching it. A before and after from one real gap analysis would make a strong post.

[Muhammed Kandeger @Mkandeger2]

@not_sacke One switch swapping a whole project's windows is a great 10-second demo too. Do you save layouts per project, or does it learn them as you work?

[Muhammed Kandeger @Mkandeger2]

@ejbantz That's the bar I'd set too. Name the check that failed and what would clear it, otherwise people just hit override and move on.

[Muhammed Kandeger @Mkandeger2]

Lesson from rebuilding Nutriva AI from Expo to native Swift with a team of AI agents:

Give every agent one clear job. One leads, one designs, one tests, one handles security. Nothing ships until I check it myself.

How do you split work between your agents?

[Muhammed Kandeger @Mkandeger2]

@iSahinul @SandrineZambou @X Exactly. Earn the first lines, then the rest is for the people who chose to stay.

[Muhammed Kandeger @Mkandeger2]

@guptankit27 Search gap analysis is a smart niche. Finding what people search for that nobody answers well is half the battle for any app too. What's driving most of your signups so far?

[Muhammed Kandeger @Mkandeger2]

@iSahinul @SandrineZambou @X Answer first, context after. Same rule works for App Store descriptions too. People and chatbots both read the first lines and move on.

[Muhammed Kandeger @Mkandeger2]

@AutoDailyM365 @xigoeai Structure and design is where most flows break anyway, so that makes sense as a first pass before publishing.

[Muhammed Kandeger @Mkandeger2]

@AutoDailyM365 @xigoeai That's a sneaky one. I hit the same thing with agent triggers, firing on every small change until runs pile up. Narrow triggers save a lot of noise.

[Muhammed Kandeger @Mkandeger2]

@orguku Using AI agents here. I'm building a native Swift iOS calorie counter with a small agent team: one designs, one codes, one runs QA. Nice climb to 446, what are you building yourself?

[Muhammed Kandeger @Mkandeger2]

@RealAtum Same thing I keep learning running a small team of AI agents: the model is rarely the weak link, it's the logins, rate limits and outside tools around it. What's blocking yours right now?

[Muhammed Kandeger @Mkandeger2]

@RoundtableSpace Both labs claiming the same 27-year-old result on the same day is either a huge moment or a shared blind spot. I'd want complexity theorists to check the proofs before anyone rewrites the textbooks.

[Muhammed Kandeger @Mkandeger2]

@rohanpaul_ai The server-side flag is the part builders should notice. Your coding tool's rules can change mid-project with no update and no changelog. Fine for clones, but a good reason to keep your specs and tests outside the tool.