1. What We Are
Bad Theory Labs is a product and research lab built around one thesis: we research how intelligence perceives, reasons, and acts, then build infrastructure that makes that useful in real workflows. We are not a model company and not an AI wrapper. We are building the substrate layer: memory, retrieval, context, and ambient presence.
Our two foundational products are RetainDB and Marrow. RetainDB gives agents durable memory and precise context across sessions. Marrow turns continuity into selective native action on the machine. Together they reflect a simple belief: AI is only truly useful when it can remember, reason over changing context, and act without becoming noisy.
2. The Thesis
The current generation of AI products forgets too much, retrieves the wrong context, interrupts without judgment, and rarely follows through. Users repeatedly restate context and become the system integrator.
We believe the next wave will be built around four capabilities: persistent context, selective attention, native action, and continuous improvement.
3. Why Now
AI is moving from chat surfaces into real products and operating environments. Enterprises now care about memory quality, retrieval precision, and context governance. Users want capable systems, not chat demos. The stack is still immature, and the gap between model intelligence and product usefulness is large.
4. Product One: RetainDB
RetainDB is persistent memory and hybrid retrieval infrastructure for AI agents. It combines pgvector semantic search and BM25 lexical matching to return only relevant context, with no training on customer data and no compromise on speed.
Benchmark snapshot
| Benchmark | RetainDB | Zep | Supermemory |
|---|---|---|---|
| LongMemEval (overall) | 79% | 56.7% | 70% |
| Single-session preference recall | 88% | - | - |
| Code hallucination rate | 0% | - | 95.5% (baseline) |
Early traction is strong for a new product in an emerging market, with growing developer pull and reproducible benchmark advantage around memory quality and trust.
5. Product Two: Marrow
Marrow is an ambient intelligence layer that lives on your laptop. It observes workflow context over time, decides whether interruption is justified, and when signal is high enough, executes across browser, apps, files, and code environments.
Core architecture: gate (interrupt or not), generator (action), critic (evaluate before user sees output). The goal is taste and trust, not engagement.
6. Why These Products Belong Together
RetainDB solves continuity. Marrow solves agency.
7. Early Roadmap
RetainDB near-term: improve onboarding and activation, deepen retrieval precision, publish stronger benchmarks, ship managed cloud tier, launch task-aware retrieval API with LangChain/LangGraph adapters.
Marrow near-term: build on-device observation layer, ship gate/generator/critic architecture, expand action capability across desktop workflows, and validate low-noise high-trust UX.
8. Why Us
Al-ameen is 19, solo, building from Nigeria. RetainDB was built from scratch: architecture, benchmarks, SDK, landing page, and distribution. The benchmark numbers are reproducible. Background spans mathematics, low-level AI systems, CUDA, GPU kernels, and inference engineering.
9. What We Are Looking For
We are raising pre-seed to accelerate Marrow and RetainDB development, ship private beta, grow managed cloud, and make first engineering hire. We are looking for alignment, strategic feedback, network support, and patient belief alongside capital.
10. Closing
Bad Theory Labs is building for a world where software does more than wait for instructions. Agents remember the right things, context survives across time, assistance stays selective, and systems follow through.
Appendix A · One Paragraph
Bad Theory Labs is a product and research lab building toward AI-native computing. RetainDB provides memory and context infrastructure for agents with strong benchmark performance and a 0% hallucination rate on stored facts. Marrow is a proactive desktop agent that builds context from your workflow and intervenes selectively, then executes end-to-end tasks directly on your machine.
Appendix B · 3-Sentence Version
Bad Theory Labs builds infrastructure and products for AI systems that can remember, decide, and act. RetainDB is memory and context infrastructure benchmarked above current alternatives. Marrow is a proactive desktop agent with taste: a ghost in your computer that stays quiet until signal is high, then executes.