There's a lot of noise around how "Ai generates buggy software". The truth is writing the code was always 20% of the time. Testing and refining was always 80%. But now as that 20% compresses 10x, we want to compress the 80% as well... That's the mistake.
Here's my simple model of the history of the software industry.
Warning: It's REALLY simple. Some of you will find it simplistic; even dumb.
But I find it helpful when thinking about new technologies and companies, especially AI!
Phase 0
In the beginning, there was no software. Everything was done manually, with some analog or digital automations like calculators and fax machines.
In those days, it was relatively easy to find applications where software was better than no software.
e.g. Spreadsheets. Word processing. Email.
Phase 1
After that, the opportunities in the software market were forever divided into two categories:
a. Existing software: To win, you had to make BETTER software AND convince customers that "better" was worth switching.
e.g. Slack instead of email. Cloud instead of on-prem. Internet instead of CD-ROM.
b. Still no software: Where things were still manual, it remained relatively easy to find applications where software was better than no software — ASSUMING it was technically possible for software to replace the manual solution.
e.g. Zoom instead of business travel. Electronic instead of manual stock exchanges. Scanner apps instead of scanners.
Phase 2
For decades, the industry marched forward. Left foot: compete with existing software. Right foot: replace a legacy process.
But there's a problem: Over time, both of these categories got HARDER to succeed in.
The existing software (a) got better and better, which made it harder and harder to compete with.
There became fewer and fewer pockets with no software (b), and in the remaining pockets, software was not capable enough to beat out the legacy solutions.
Until...
Phase 3
In the past few years, advances in AI have brought dramatic new capabilities to software.
All of a sudden, it became EASIER to succeed in both categories:
First, AI software has the potential to be WAY better than existing software (a).
e.g. AI notetakers like Fathom over call basic transcription. LLM chatbots over search engines. Industrial control agents like Phaidra over hard-coded control systems.
And finally, AI now makes it possible to replace some of the legacy manual processes (b) with software for the FIRST time.
This opens up MASSIVE new areas of opportunity!
e.g. Drug development. Marketing. Medical diagnosis. Material science. These important parts of the world economy were — until very recently — still largely manual, time-intensive processes that AI is improving and accelerating.
At @charactervc, we invest in both categories — but it's important to understand the competitive dynamics before (and after) making an investment.
Now, the trillion-dollar question: What happens next?
After the current flurry of activity, will we end up stagnated in something like Phase 2 all over again? Or will AI development keep pace — moving fast like a run instead of a march?
Only time will tell — and it's one of the reasons it's so exciting to be working in this field today.
@Daveeys21 Gran oportunidad de aprendizaje, mantener la concentración y controlar los nervios no se aprende en otros espacios.
Tienen el talento y la técnica, sigan acumulando experiencias con la frente en alto.