Memla / Acquisition Preview
Memla turns accepted coding work into transferable intelligence.
Instead of only retrieving similar text, Memla stores roles, constraints, transmutations, and verification rituals from accepted work, then uses that structure to improve the same model on future coding tasks and on new repos.
Headline proof / same model, unseen second repo
0.1667->0.9167
Raw file recall to Memla-assisted file recall on unseen tasks, with no target-repo modification.
Public curriculum
6 / 7
Public repos reached holdout across web app, backend API, auth/security, and CLI/tooling families.
Transfer unit
Transmutations
Memla stores the constraint trade behind a successful fix, not just the file path that changed.
Why buyers care
This converts expensive frontier-model usage into owned repo intelligence that compounds instead of evaporating between tasks.
Proof Layers
The headline is the same-model second-repo jump. The support layers show that the bootstrap path also survives public seeded runs and a broader multi-family curriculum.
Recall graph
Baseline / rawMemla
Why This Is More Than RAG
The difference is not "more context." The difference is that Memla stores reusable problem structure and then maps it onto the next repo.
Naive retrievalPulls similar text chunks and hopes the model can reconstruct the workflow.
MemlaStores role targets, constraint tags, transmutations, and verification rituals from accepted work.
Failure modeLooks smart inside one prompt, then relearns too much on the next repo.
AdvantageEnters new repos with a stronger first structural prior and compounds faster after foothold.
What Memla Actually Stores
Accepted trace
|- likely files
|- commands / tests
|- workflow steps
|- repo family
\- acceptance outcome
Transmutation layer
|- role targets
|- constraint tags
|- transmutations
\- verification rituals
Transmutation Bank
These are the recurring constraint trades Memla now reuses across public repo families.
Selected Proof Cases
These examples show how Memla steers the same model toward the local repo shape instead of generic architecture guesses.
Multi-Family Support
The new curriculum rerun matters because it shows seeded support outside the original repo cluster and across more than one architecture family.
Completed public holdouts
Async diligence friendly
Ready to send, not just ready to explain.
Memla already has the full diligence packet, proof table, and frozen reports. If this is relevant to your roadmap, the fastest next step is an async diligence exchange on email.