Building Panzoto - an Android app that turns your voice recordings into searchable transcripts using Whisper AI.
Looking for beta testers! Free credits available.
DM me if you want to try it out and give feedback.
#BuildInPublic#AI#ProductivityApp
I have recently been trying to design a product that help mid-level managers to make decisions faster and more consistent. This first issue with product is that the manager themselves doesn't necessarily need the product. Because people already know how they would make decisions
With all these easy to make #NeuralNetworks, it's essential to customize your model. Otherwise, it's easy to compromise your security if people know what your does. #security https://t.co/86MW5MxIS3
Tired of spending so much time making decisions? Worried about automatic decisions result in decisions that's not you? Try this decision making tool that customize your decision on a whole new scenario, based on your past decisions. https://t.co/dyNi4z0ZVa
FTC is trying to block the Microsoft/Activision deal. I'm in favor of it for all the wrong reasons. I don't want Blizzard, or what's left of it, to become EA. It's a losing fight, but my childhood is trying to hold on to the last few pieces that's still there.
After each interview, make sure to debrief and compare notes. If there is a pain point that both person agree to, then it's obvious which direction we should go. But there is often the notetaker observed some detailed that the interviewer might have missed.
Recording the interview can also be beneficial. Ask for permission, and state it will greatly benefit you later. Rewatch the recording can help you pick out visual cues you missed in the first run. Notetaker should observe the person's behavior, but also body language.
This might be counterintuitive. When interviewing about the potential usefulness of a product, try to not pitch the idea. Rather, ask the person to describe the problem you are trying to solve and how it affect the person's life.
When interviewing, it's often helpful to have a two person team. While one person ask the questions, the other person observe reactions and take notes. The founder should be the observer if possible. This is to avoid the problem of interviewee tell what you want to hear.
In those interviews, ask for: 1) the current solution, 2) possible alternative solutions, and 3) pitfalls of the current solution. Keep in mind that it's often the case there are no good solutions exist. That's a good thing and it indicate a need.
As a rule of thumb, interview 5 people when doing market research. The exception is when you have multiple customer segments. Then it's probably good to interview 10 to at least get an idea of the plan.
When designing interview questions for marketing research, it's helpful to write down the questions and what to say then the interviewee starts to deviate from the questions. Unless what they say spark new ideas about your product, it's better to focus on your questions.
@johnbknechtel I see this as a missed opportunity for meta to not capitalize on their latest model. OpenAI makes clear offers for using there LLMs with per usage price. meta doesn't seems to be interested at all. Prompt engineering is here to stay. Companies need an easy way to customize.
LLaMA and models derived from it have not been cleared by Meta for commercialization permissions. As useful as it is, it's not a viable product option.
In the idea validation phase, write an interview guide and talk to potential customers. The interview guide needs to be written before the interview. Write down, 1) they type of info you wish to learn, 2) intro to your company, 3) customer segment, 4) problems with each segment.