For anyone interest in ai of course.
Well I may have some venture capital interest now. And typically I had a major problem with my system which I knew was coming.
As launch took longer than expected I didn't have much data saved into my system. Then the LLM model I was using ended. The new updated models never worked as well. I was only using Claude for 10% of my system unlike others who use it 100% but it was enough to cause me a real headache.
Then I had a bright idea - which cost me 80 hours of intense stressful work.
I decided to build a system linking ai - any ai - to grounded structured databases. It sounds really easy - but it nearly broke me. Seriously the hardest thing I have ever done. And I finally cracked it. I am using 80% of my old system but with this new layer.
I didn't realise I had independently built a neurosymbolic ai system that actually works. They are the next big thing in Ai but nothing really working outside research labs.
So now my system has gone from 90% accuracy to 100% accuracy which is vital for things like customs or any other regulated industry.
I now have plugins for each country's official tariff data etc. So my system can become truly international. But without VC backing it could take me a long time to get where I want to be.
I have never liked or trusted British Venture Capital. Basically if you don't have the right background or your face doesn't fit they don't want to know! Hopefully I have met a chap who can change that - we shall see.
Well I may have some venture capital interest now. And typically I had a major problem with my system which I knew was coming.
As launch took longer than expected I didn't have much data saved into my system. Then the LLM model I was using ended. The new updated models never worked as well. I was only using Claude for 10% of my system unlike others who use it 100% but it was enough to cause me a real headache.
Then I had a bright idea - which cost me 80 hours of intense stressful work.
I decided to build a system linking ai - any ai - to grounded structured databases. It sounds really easy - but it nearly broke me. Seriously the hardest thing I have ever done. And I finally cracked it. I am using 80% of my old system but with this new layer.
I didn't realise I had independently built a neurosymbolic ai system that actually works. They are the next big thing in Ai but nothing really working outside research labs.
So now my system has gone from 90% accuracy to 100% accuracy which is vital for things like customs or any other regulated industry.
I now have plugins for each country's official tariff data etc. So my system can become truly international. But without VC backing it could take me a long time to get where I want to be.
I have never liked or trusted British Venture Capital. Basically if you don't have the right background or your face doesn't fit they don't want to know! Hopefully I have met a chap who can change that - we shall see.
