Many AI startups building agents and getting funding have a tough road ahead if they want to go public (IPO). More likely, they'll get bought by big tech companies.

As a software developer using tools like ChatGPT, GitHub Copilot, and Cursor, I see that today’s chatbots struggle with basic tasks and have pretty high error rates, around 5% or more. This makes them unreliable, especially when they need to work with multiple software pieces. Creating effective agents involves many steps, with chatbots being just one part.

Even though these technologies might get better in five years, startups can't currently compete with well-funded companies like OpenAI or Anthropic. Using models like GPT-3, startups are unlikely to reduce error rates enough to succeed in the near term.

🇧🇷🇺🇸-  Book: Making Things Think: https://holloway.com/mtt. Investor in Wander, Carry, Footprint, Merkle Manufactory (Farcaster), Dynamic, Paragraph.

Not as sure about that. 5 years is an eternity. There are good agent based systems being built that solve real problems today. Most of them are based on Graph RAG which helps them have more relevant answers and agents are being built to integrate with enterprise systems via tools. The last piece that is missing is intra-agent communication and that's not going to take 5 years to happen.