👤 Who I am: Richard Danks — Hinckley, Leicestershire (UK Midlands) 🛠️ What I actually do: I run DDM Solutions Ltd as a fractional CTO / Chief AI Officer, and I'm a partner at Assured Velocity, a transformation consultancy. Practically: I'm the person mid-market firms bring in when they've got a transformation programme or an AI ambition and no senior technologist to own it. Six years of agency-side marketing before the 25 years in enterprise IT, so I sit in the odd overlap between marketing and architecture — which is why I'm in this group. 🤖 What I'm building with AI right now: Two builds and one obsession. The builds: a signals-and-prospecting engine - a multi signal registry watching for buying triggers, enriching against Companies House, running a doer/verifier agent pattern so nothing goes out on a hallucinated fact. AEO/GEO work: rebuilding client entity footprints (schema, Wikidata, sameAs graphs) so they get cited by LLMs, not just ranked by Google. The obsession is what sits underneath both: Extending AI memory beyond flat retrieval, using human cognition as the reference design. So, typed memory rather than one blob, graph structure over entities and relationships so a model can traverse instead of just match, ML/NLP deciding what's worth keeping in the first place, and forgetting treated as a feature rather than a failure. The part I find most interesting is where you shouldn't copy biology. Machines get perfect recall and no decay for free, so the goal isn't to simulate a brain, it's to take the bits evolution got right and drop the bits that were just constraints. Entity work for search and memory design for agents turn out to be closer to the same problem than I expected, both are about giving a machine a durable, structured idea of who someone is. 🎯 What I'm looking for connection-wise: 🧰 Trading workflows & systems I'll happily open up my prospecting stack and agent patterns if you'll show me yours 📬 Best way to reach me: DM here, or LinkedIn — linkedin.com/in/richarddanks

👤 Who I am: Richard Danks — Hinckley, Leicestershire (UK Midlands) 🛠️ What I actually do: I run DDM Solutions Ltd as a fractional CTO / Chief AI Officer, and I'm a partner at Assured Velocity, a transformation consultancy. Practically: I'm the person mid-market firms bring in when they've got a transformation programme or an AI ambition and no senior technologist to own it. Six years of agency-side marketing before the 25 years in enterprise IT, so I sit in the odd overlap between marketing and architecture — which is why I'm in this group. 🤖 What I'm building with AI right now: Two builds and one obsession. The builds: a signals-and-prospecting engine - a multi signal registry watching for buying triggers, enriching against Companies House, running a doer/verifier agent pattern so nothing goes out on a hallucinated fact. AEO/GEO work: rebuilding client entity footprints (schema, Wikidata, sameAs graphs) so they get cited by LLMs, not just ranked by Google. The obsession is what sits underneath both: Extending AI memory beyond flat retrieval, using human cognition as the reference design. So, typed memory rather than one blob, graph structure over entities and relationships so a model can traverse instead of just match, ML/NLP deciding what's worth keeping in the first place, and forgetting treated as a feature rather than a failure. The part I find most interesting is where you shouldn't copy biology. Machines get perfect recall and no decay for free, so the goal isn't to simulate a brain, it's to take the bits evolution got right and drop the bits that were just constraints. Entity work for search and memory design for agents turn out to be closer to the same problem than I expected, both are about giving a machine a durable, structured idea of who someone is. 🎯 What I'm looking for connection-wise: 🧰 Trading workflows & systems I'll happily open up my prospecting stack and agent patterns if you'll show me yours 📬 Best way to reach me: DM here, or LinkedIn — linkedin.com/in/richarddanks

👤 Who I am: Richard Danks — Hinckley, Leicestershire (UK Midlands) 🛠️ What I actually do: I run DDM Solutions Ltd as a fractional CTO / Chief AI Officer, and I'm a partner at Assured Velocity, a transformation consultancy. Practically: I'm the person mid-market firms bring in when they've got a transformation programme or an AI ambition and no senior technologist to own it. Six years of agency-side marketing before the 25 years in enterprise IT, so I sit in the odd overlap between marketing and architecture — which is why I'm in this group. 🤖 What I'm building with AI right now: Two builds and one obsession. The builds: a signals-and-prospecting engine - a multi signal registry watching for buying triggers, enriching against Companies House, running a doer/verifier agent pattern so nothing goes out on a hallucinated fact. AEO/GEO work: rebuilding client entity footprints (schema, Wikidata, sameAs graphs) so they get cited by LLMs, not just ranked by Google. The obsession is what sits underneath both: Extending AI memory beyond flat retrieval, using human cognition as the reference design. So, typed memory rather than one blob, graph structure over entities and relationships so a model can traverse instead of just match, ML/NLP deciding what's worth keeping in the first place, and forgetting treated as a feature rather than a failure. The part I find most interesting is where you shouldn't copy biology. Machines get perfect recall and no decay for free, so the goal isn't to simulate a brain, it's to take the bits evolution got right and drop the bits that were just constraints. Entity work for search and memory design for agents turn out to be closer to the same problem than I expected, both are about giving a machine a durable, structured idea of who someone is. 🎯 What I'm looking for connection-wise: 🧰 Trading workflows & systems I'll happily open up my prospecting stack and agent patterns if you'll show me yours 📬 Best way to reach me: DM here, or LinkedIn — linkedin.com/in/richarddanks

@Pascal Pollack the Signals Registry is 80 % code pulling from APIs and then cross matching signals from different sources against a company before the LLM is used to refine a report that covers whats been found. There is a QA check against rules. There's also an ML element in there thats improving as I get more data. (its checking against segments of 8 years of accounts to prove various thesis like X happens plus Y = company in administration). Thanks for the info on the doer verifier pairing 'll do some tests on my stuff.