My First Interview Question: What Side Projects Are You Working On? | The Daily Hello Ep 006

My first interview question isn't about your past. It's "what side projects are you working on?" Then: how many mistakes have you made, and how big is your imagination? This is the Oct 6 Daily Hello: three of us building Model Deployment Company in public, on how we'd hire when the technology itself is only 2 or 3 years old. Plus: a model built to hallucinate on purpose. In this episode: 00:00 Cold open: "what side projects are you working on?" 00:10 What's broken in today's AI workflows 01:16 What if a model is built to hallucinate? 03:29 Most people use AI to be entertained 07:27 Too many problems: how to pick one 08:42 Newton, gravity and asking a new question 11:10 10 problems, pick 5, start with 2 13:36 Burning credits: open models, Cursor and Claude 16:41 Re-invent yourself: imagination over textbook answers 18:35 One mistake a day, 100 problems in 90 days 19:25 Screen share: Gemini on LLM security 21:23 How would you hire in the AI era? 25:59 Experience is the speed of problem solving 27:44 You can't have 8 years in a 2-year-old technology 30:06 My first interview question: side projects 31:12 Second: how many mistakes. Third: imagination 34:30 Real intent won't let you sleep 36:55 Who gets paid? Whoever finds the right problem 37:16 Cursor in Slack, Bugbot and burnt credits What we talk about: • Hiring in the AI era: side projects (no one is paying you, no manager), mistakes (the successes are one, two or three; the mistakes are 50) and imagination • You can't get 8 years of experience in a technology that's 2 or 3 years old, so experience becomes the speed of problem solving • The opposite approach: a maximally hallucinating model, because creative work (images, video, new ideas) needs hallucination • Most people spend 1–2 hours a day on useful AI and 13–14 hours entertaining themselves • One mistake a day: try 100 problems in 90 days instead of waiting for the one right problem • Rithvik's screen share: asking Gemini about LLM security • Open models vs. Cursor and Claude subscriptions, and where our credits went (Bugbot) Hosts: Akhil Kolla, Rithvik (Sai) Panchumarthi, Gopi Model Deployment Company — building a company in public. #BuildInPublic #AI #ModelDeployment