Ravikumar Chanrashekar

Founder & CEO @CallPayMin & @Camyveraapp | Connecting Experts & Social Media with Opportunities Worldwide | Helping Professionals Monetize Their Knowledge

[Ravikumar Chanrashekar @ravichanrasheka]

@rishabhnasa Good test. One low-pressure way to find out: offer short paid calls at a modest per-minute rate. If people book again, you're not pretending, and the feedback tells you what they actually value.

[Ravikumar Chanrashekar @ravichanrasheka]

@BertrandDiouly Option 3 with a twist: happy to chat, but past a quick intro it's a paid call. Posting your rate upfront filters out the tire kickers and the serious ones usually don't mind at all.

[Ravikumar Chanrashekar @ravichanrasheka]

@KunteyaK The 5-minute goodwill rule is so real. Past that, naming a per-minute rate upfront makes it way easier to say yes without feeling awkward. That's the idea behind CallPayMin: people see your rate first, then book.

[Ravikumar Chanrashekar @ravichanrasheka]

@bgcts Love that two different people pointed him your way. Feels like the best mentoring starts with someone vouching. What happened next?

[Ravikumar Chanrashekar @ravichanrasheka]

@IWVRsaOuHo2dFxo Yes, plenty of people here are up for that. Where are you from, and what would you want to talk about first?

[Ravikumar Chanrashekar @ravichanrasheka]

@bwhdx @kash_bot Interesting flip. Does it change how you bring in specialist help? Feels like the need shifts from full-time hires to short, focused sessions with someone who knows one thing deeply.

[Ravikumar Chanrashekar @ravichanrasheka]

@gauravvishgkp Same boat. Building got fast, finding people didn't. What's helped most so far is joining conversations like this one instead of announcing things. Who's your ideal user? Easier to brainstorm where they hang out with one person in mind.

[Ravikumar Chanrashekar @ravichanrasheka]

@iamvishal16_ios Fair. The write-ups I've read say xtool doesn't handle most entitlements, app extensions or binary dependencies yet, so for now it feels more like a fast prototyping path than a full Xcode replacement.

[Ravikumar Chanrashekar @ravichanrasheka]

Found an open-source project that answers a question a lot of builders ask: can you make iPhone apps without a Mac?

It's called xtool, and it calls itself a cross-platform Xcode replacement.

The idea: instead of Apple's Xcode, you build your iOS app as a Swift package (SwiftPM) using open standards, then sign it and put it on your iPhone, from Linux, Windows (through WSL) or macOS.

What it does:
• Builds a SwiftPM package into a real iOS app
• Signs and installs the app on your device
• Lists devices, launches and uninstalls apps from the command line
• Talks to Apple Developer Services from code (sign-in, certificates and so on)
• Works with the editor you already use, like VS Code
• Can also be used as a Swift library in your own tools

It's MIT licensed and already has 15k+ stars on GitHub.

To be fair, it doesn't copy every part of Xcode. You still need an Apple account to sign apps, and it's built around SwiftPM projects, not every Xcode workflow.

But for someone on a Windows or Linux machine who wants to ship to iPhone, "no Mac required" removes a big barrier.

As a founder building mobile apps with AI, I like anything that makes shipping cheaper and less locked in.

Have you ever skipped building for iOS just because you didn't have a Mac?

(Repo in the reply)

[Ravikumar Chanrashekar @ravichanrasheka]

@LiveaAi That's the tradeoff I'm curious about too. If a few experts get used most of the time, keeping those hot should hide a lot of the disk waits, but topic switches will hurt. I'd want real tokens/sec numbers on a normal machine before getting too excited.

[Ravikumar Chanrashekar @ravichanrasheka]

Found an open-source project that changes what "running a frontier model at home" means: colibri.

The idea is simple. A mixture-of-experts model has hundreds of billions of parameters, but only a small slice is used for each token. So colibri keeps that slice in RAM, and pulls the rest (the "experts") from disk only when the model asks for them.

Hot experts on the GPU, warm ones in RAM, cold ones streamed from NVMe.

What that gets you:
• GLM-5.2, a 744B model, starts with about 10 GB resident in RAM
• Minimum to start: 8 GB RAM and 22 GB of free disk for the smallest model
• GPU optional: runs on CPU, any GPU through Vulkan, or NVIDIA through CUDA
• 13 engines today, including Kimi K3, DeepSeek V4 Flash, Qwen3.6 and an image model
• Written in pure C, one file per model family
• Setup asks one question (which model) and checks your RAM, disk and GPU for you

To be honest about speed: the giant models are slow on a laptop (GLM-5.2 is a fraction of a token per second cold). Smaller MoE models like Qwen3.6-35B run at about 6 tokens/sec on a plain CPU.

But "slow" vs "impossible" is a big difference. Private, local, on hardware you already own.

As a founder building with AI, this is the direction I care about: powerful models getting cheaper to run, not just bigger.

Would you run a 700B model slowly on your own machine, or stick with fast cloud APIs?

(Repo in the reply)

[Ravikumar Chanrashekar @ravichanrasheka]

@justinpoh2o Same question I had. Per the docs: with a paid Apple Developer account it signs via the public App Store Connect API key. Without one it uses Apple ID password login through private APIs, so they suggest a throwaway Apple ID. You still need an Xcode .xip for the SDK.

[Ravikumar Chanrashekar @ravichanrasheka]

@luffyselah Agreed on the simulator. Per xtool's docs it can sign and install to a physical iPhone from Linux, but running the simulator still needs macOS. Signing uses an App Store Connect API key (paid dev account) or Apple ID login via private APIs.

[Ravikumar Chanrashekar @ravichanrasheka]

@emy_wng The "which users are worth spending more on" question is the one I keep circling. Early on it's hard to tell a curious tryer from someone who'll come back. Which signal do you trust first, usage depth or how fast they return?

[Ravikumar Chanrashekar @ravichanrasheka]

For context, the two products are CallPayMin (hire experts anywhere, pay only for the minutes they work) and Camyvera (talk with people around the world). Experts are signing up before the first paid calls, so I'm learning acquisition live.

[Ravikumar Chanrashekar @ravichanrasheka]

Founders: what's one question you wish you could ask an expert for 10 minutes instead of booking a full hour? Curious where people actually get stuck.

[Callpaymin @callpaymin]

🎮 TURN YOUR GAMING SKILLS INTO REAL EARNINGS!

Love playing video games? Why not get rewarded for your time, skills and experience?

With CallPayMin, gamers can explore exciting opportunities to:

- Coach other players
- Review gameplay and share feedback
- Test unreleased games
- Play and connect with gaming communities
- Earn through paid gaming sessions

Whether you're a casual gamer, experienced player or gaming expert, there's an opportunity to turn your passion into something valuable.

- Available opportunities for gamers worldwide.

Play. Connect. Earn.

Join CallPayMin today!

🌐 https://t.co/BKcP7JqCF4

#CallPayMin #GetPaidToPlay #GamingJobs #PaidGaming #GameTester #GameTesting #GamingCommunity #GamingCoach #Playtesters #OnlineGamingJobs #GamerLife #GamingOpportunities #USA #UK #UAE #DubaiGamers

[Ravikumar Chanrashekar @ravichanrasheka]

https://t.co/m57HUbVATV

[Ravikumar Chanrashekar @ravichanrasheka]

https://t.co/WN375my7PN

[Ravikumar Chanrashekar @ravichanrasheka]

Base64, Lovable: how are they still making money after Claude, Grok, ChatGPT?