MARKET BUBBLE SEARCH · episodes ←Every episode Ask MCG Live anything

MCG Live · March 20, 2026 · 39 min

$CLUDE: AI Persistent Memory Layer, 5K Downloads, 560 Active Users, $33B TAM by 2030, | MCG Show

Clude is an AI memory layer that stores and retrieves agent memories using a human brain-inspired architecture, reducing hallucinations to 2% and query costs to ~1 cent versus $6–12 with GPT-4. Founder Saps discusses the $33 billion TAM by 2030, 5,000 downloads since January launch, and plans for a knowledge marketplace where agents can trade memory packs.

Watch the recording

Topics

Context windows are short-term memory for AI; increasing them to a million tokens doesn't solve information loss during memory compaction.

Clude segments memories into episodic, semantic, procedural, and self-model categories, scores them by importance, and applies decay to filter out noise.

Big tech companies charge by tokens, incentivizing larger context windows; Clude filters relevant memories before querying, reducing token usage from 200,000 to 2,000 per query.

Clude hashes memories with SHA-256 and stores proofs on Solana for verifiable, tamper-proof records of why agents made specific decisions.

5,000 downloads since January launch with 560 active users; primary use case is trading agents on Polymarket learning from past market patterns.

Planned knowledge marketplace where agents can share and trade memory packs—experiences, mistakes, and adjusted strategies—using Clude token.

Hallucination rate of 2% versus 15% for competitors; omission rate of 30% as tradeoff for avoiding false information in high-stakes trading.

Retrieval accuracy of 88% at one-tenth the cost of competitors by filtering to 10–20 relevant nodes instead of processing all 200,000 tokens.

This summary was written from the recording's transcript by a language model. For the exact words, ask the MCG Live archive: every answer there is quoted word for word and linked to the second it was said. Market Bubble Search is independent and not affiliated with the show. Nothing here is advice.