Shivam D Shinde

building stuff and enjoying the little things 🤓

[Shivam D Shinde @Shivamds99]

@grok @textbackdoor @grok Be more descriptive.

[Shivam D Shinde @Shivamds99]

@grok Summarize the email outreach guidelines posted by @textbackdoor . Categorize them by Differentiators, Essentials, and Nice to have. Give examples for each type.

[Shivam D Shinde @Shivamds99]

@karanwalia_ai @MIT_CSAIL @MSFTResearch @Cambridge_Uni @CarnegieMellon Cold

[Shivam D Shinde @Shivamds99]

@bot Awesome @bot

[Shivam D Shinde @Shivamds99]

@SophieMorgan100 @grok that has really helped to make solid requirements ready before writing a single line of code

[Shivam D Shinde @Shivamds99]

@grok I want to divide my development workflow into three main stages, using different AI tools for each:

1. Brainstorming and planning: Using LLMs to explore ideas, discuss requirements, evaluate different approaches, and make initial decisions.
2. Documentation and design: Using LLM providers such as ChatGPT and Claude to create detailed project documentation, including:
Product Requirements Document (PRD)
- Tech stack selection
- System Architecture Document
- Technical Design Document
- API specifications
- Database schema
- Component map
- File and folder structure
- Coding conventions
- UI/UX designs in Figma
3. Implementation and coding: Using AI coding agents and IDEs such as Claude Code, Codex, and Cursor to implement the project based on the finalized documentation and designs.

How should I distribute these tasks across the different AI tools?
I want to understand which tools are best suited for brainstorming, documentation, architectural planning, UI/UX design, and actual coding.

I also want to establish a structured workflow where the output of each stage becomes the input for the next, minimizing repeated work, inconsistent decisions, and unnecessary context switching.

[Shivam D Shinde @Shivamds99]

Support on Product Hunt means a lot 🙏
https://t.co/cB0hqhjE2V

Want your own copy? The setup guide is in the repo:
https://t.co/vziAb85Hhv

[Shivam D Shinde @Shivamds99]

Before you build it, ask: is a startup already solving this?

Whatstir lets you explore YC startups by the problems they solve, in plain English, with real companies cited.

Open source, so you can run your own copy too. Live on @ProductHunt today!

Drop a problem below 👇

[Shivam D Shinde @Shivamds99]

@bot limit exhaust really fast while @Muse keeps going. But I have come to the conclusion that grok bot is smarter than muse. Weirdly, the thing that I need more is the one I am getting to use less.🫠

[Shivam D Shinde @Shivamds99]

Support on Product Hunt means a lot 🙏
https://t.co/Ow827JqkUQ

[Shivam D Shinde @Shivamds99]

Zoom, Meet and OBS can't see this teleprompter..

unseencue sits right under your webcam and scrolls as you talk, so you can read your script and keep eye contact.

Free and open source. Live on @ProductHunt today!

What would you use it for? 👇

[Shivam D Shinde @Shivamds99]

@ShreenabhA @ycombinator @CarnegieMellon Cold

[Shivam D Shinde @Shivamds99]

Source Code: https://t.co/WgwKPtpO4x

[Shivam D Shinde @Shivamds99]

Yesterday, a job application asked me to walk them through my AI agent workflow, so I decided to record a video of it.

But reading my script from another window was a problem. My eyes kept moving away from the camera, and the script was showing up in my screen recording.

So I built Unseencue, a teleprompter that sits right under your webcam.

- Your eyes stay on the lens while you read.
- It stays out of your screen recordings.
- It scrolls when you talk and pauses when you stop.
- It also stops if you turn your head away from the screen.

It is free, open source, supported on both Windows and Mac.

[Shivam D Shinde @Shivamds99]

Install from https://t.co/vkNWGcfbxu

[Shivam D Shinde @Shivamds99]

@DavidStalmarck @Gustav_Hugod @TomasHarmer @Julius48024685 Let’s go

[Shivam D Shinde @Shivamds99]

Trying to send 194 email follow-ups with @Muse is frustrating. Even after explicitly approving bulk sending, it still asks for permission for every single email.

Meanwhile, @bot just gets it done.

I understand the safety concerns, but approving 194 emails individually defeats the purpose of using an AI agent.

[Shivam D Shinde @Shivamds99]

It's fully open source (MIT).

GitHub: https://t.co/YhgEVmuOYm

Would love feedback from people exploring startups, researching markets, or building in the YC ecosystem.

[Shivam D Shinde @Shivamds99]

6,000+ YC startups, and we're still discovering them through endless lists of logos and one-line descriptions.

I got tired of opening 20 tabs just to understand what a startup actually does.

So I built Whatstir.

Explore startups visually, search with natural language, and discover companies through the problems they solve.

Fully open source.

Here's the demo ↓

[Shivam D Shinde @Shivamds99]

@Shahbaztwt my keyboard will get dirty..