If you’re new here, welcome to THIRD’s dealflow newsletter. Each week, we’ll highlight one opportunity featuring global startups raising at the angel, pre-seed and seed level 👼

If you want to get in touch with the founders, just reply back to this email and we’ll introduce you.

You can view past deals here - just make sure to enter your email for access and be a paid subscriber!

The Problem 🚨

Today’s AI agents rely on RAG pipelines and vector databases to retrieve information. But retrieval isn’t memory.

As agents become responsible for handling customer conversations, legal documents, healthcare records and financial workflows, developers are running into the same issues:

Vector search returns similar information, not exact facts.
Large context windows still aren’t enough for long-term memory.
Memory infrastructure is expensive, slow to maintain and introduces privacy and reliability risks.

The result? AI agents that still hallucinate, forget previous interactions and require complex backend infrastructure to function.

The Solution💡

Memvid is building a new memory layer for AI.

Instead of relying on vector databases and RAG pipelines, Memvid packages everything an AI agent needs into a single portable memory file.

That means:

Persistent long-term memory in one portable file
Instant retrieval without vector databases or servers
Time-aware memory that lets agents recall information exactly as it existed at a specific point in time
Up to 93% lower infrastructure costs while simplifying deployment

The Team 👥

Memvid is founded by repeat entrepreneurs who have spent years building AI products and infrastructure.

Mohamed (CEO) previously led marketing for a media company acquired by Morning Brew before co-founding a B2B growth agency helping AI startups land enterprise customers.
Saleban (CTO) holds dual Master’s degrees in Computer Science and Machine Learning, with more than a decade building AI systems across enterprise and consumer applications.

The founders first experienced the problem while building their previous healthtech company, where AI systems needed to securely remember precise patient information without relying on complex memory infrastructure.

Traction 📈

Despite launching recently, Memvid has already built strong developer momentum.

40,000+ developer users
12,000+ GitHub stars
#1 Trending GitHub repository
Paying customers today
Enterprise pipeline across fintech, legal and private equity
$20K ARR with a target of $1M ARR by year-end

Independent benchmarks also show Memvid outperforming leading memory systems, including Mem0 and OpenAI’s memory implementations on the LOCOMO benchmark.

Current Raise 💰

Pre-Seed: $800,000

The company is raising to accelerate product development, launch Memvid Cloud, expand enterprise security features, grow to 250 enterprise customers and reach $1M ARR.

Pitch Deck 📄

Upgrade to access the full deck.

Please respond directly to this email to request an introduction to the Memvid team.

Investing in early stage startups is inherently risky. Do not invest more than you can afford to lose. No communication through this newsletter or in any other medium should be construed as a recommendation for any security offered on or off this site. Investment opportunities offered by companies and founders you may connect with on this newsletter are private placements of securities that are not publicly traded, involve a high degree of risk, may lose value, are subject to holding period requirements and are intended for investors who do not need a liquid investment.

Every AI Agent Needs Memory: 40K Developers…