Minimum Viable - Oct. 8, 2025

IDEA: According to a recent article by TechRadar, AI users are not impressed with the latest release of ChatGPT. Did Sam Altman and OpenAI overpromise and underdeliver with GPT-5? Maybe. But as the product gets more complex and capable, the user’s input becomes more and more important. After all, the chatbot is only as good as the prompt it is given. There needs to be a way for non-technical users to consult with AI experts to help refine their prompts to get the most out of AI chats.

PRODUCT: Ah, Stack Overflow. A coder’s best friend. The concept is simple. Users post the problem they are having and software engineering experts are standing by to answer. Now apply that to AI. Create a forum style platform where users post the problem they are having, the prompt they are using, and their desired outcome. Contract AI experts to answer questions and suggest which platform is the best for their use case. Users would then up or downvote answers based on their experience. At first, the questions would be primarily answered by your staff, but as time goes on, other experts or community members could be upgraded to assist with prompting.

MVP: Basic online forum platform with voting system for answer correctness.

REVENUE MODEL: Advertiser based with paid subscription for premium plans.

EXIT STRATEGY: Sell to an AI giant like OpenAI or Anthropic for 10x revenue.

What do you think of the idea? Send us your feedback here.

Publication: TechCrunch

Link: Startup Battlefield company ÄIO invented a method to make edible fat from ag waste like sawdust

Summary: ÄIO (pronounced “aio”), a startup selected for TechCrunch’s Startup Battlefield competition, has developed a novel process that converts agricultural waste, such as sawdust, into edible fats. The founders aim to address sustainability and food security issues by producing fat ingredients for food applications, like cooking oil, without relying on conventional oil-producing crops. Their technology reportedly uses metabolic engineering to turn biomass into lipid precursors, then refines these into fats usable in the food industry.

Spin: This is a huge deal. Not only is this helping the food crisis many places around the world are experiencing, but this is also an incredible use of agricultural byproducts that usually go to waste. Some very well-known brands were started from byproducts. Kingsford Charcoal and Post-it Notes are just a few examples, but if you’re looking to start a company with very little raw material cost… look at byproducts that are going to waste.

Publication: TechCrunch

Link: A 19-year-old nabs backing from Google execs for his AI memory startup, Supermemory

Summary: Nineteen-year-old founder Dhravya Shah is building the startup Supermemory. It is an AI memory app designed to help AI applications retain and access long-term contextual knowledge. He started the project during a weekly build challenge he set for himself, and it evolved into a memory API that ingests unstructured data like chats, documents, files, emails, and links then surfaces relevant “memories” or insights to applications. Supermemory builds a knowledge graph to help apps query across months’ worth of data and support multiple inputs.

Spin: Have you every been chatting with ChatGPT and it all of a sudden, it seems to forget key details about your conversation? Well, this is because many AI models are great for a single session but once they get to a certain point, gaps begin to form in the memory. The gaps become larger the longer the session goes. Shah’s company is aiming to fix that allowing for much more advanced learnings and insights.

Publication: Wired

Link: Vibe Coding Is the New Open Source—in the Worst Way Possible

Summary: Developers are increasingly using AI to generate boilerplate or reusable code by “vibe coding” instead of writing everything from scratch. This is a parallel to how open source code became standard. But this shift carries serious security risks. AI models are often trained on old, vulnerable, or low-quality code, meaning those vulnerabilities can be replicated in new software. Also, generated code may not account for project-specific context, increasing chances of misconfiguration or hidden flaws.

Researchers warn of several challenges: tracing ownership of AI-generated code is harder than with open source; accountability and auditing trails are limited or absent. Plus, teams often lack policies or approved tools around "vibe coding" usage, even though many organizations report a large share of their code base now originates from AI.

Spin: As a software developer myself, I am not worried about “vibe coding” messing with security. In my experience, to get code into production environments, you had to create a pull request where it is subject to scrutiny on conciseness, correctness, and standard adherence. If a companies programming standards are strong this shouldn’t be an issue. The part that really annoyed me was the part about models being trained on low-quality code. Let’s not pretend that developers don’t copy and paste from Stack Overflow already.

We learn a lot from our successes but we learn a hell of a lot more from our failures. And in the startup world, we have plenty of people who have failed before us. One person that we can learn a lot from is Elizabeth Holmes, another member of the 30 under 30 prison club (remember Charlie Javice from Monday’s issue?). Holmes shot for the moon, but ended up flying directly into the sun. Her revolutionary blood testing company, Theranos, was shut down in 2018 after a settlement with SEC barred the young CEO from leading a company for 10 years. I could go on and on about this this, but the ABC News podcast, The Dropout, tells this story perfectly. From the rise and fall of Elizabeth Holmes to the trial and sentencing, this podcast is a must-listen.

Listen on Spotify.

IDEA: Getting sleepy throughout the day is part of being an adult. Especially with a newborn daughter who won’t sleep through the night, I would actually consider using this. Imagine you’re on a long road trip and you’ve been driving for hours. You could stop and get your third Red Bull or you could get some shut eye before getting back on the road. Now which sounds safer?

Send us an email. Let us know what you’re building, what you thought about today’s newsletter, or whatever else is on your mind.

Follow us on TikTok for a behind the scenes look at building Minimum Viable.