Blog
The why behind the what.
Engineering notes, research, opinion. We write to share what we've learned — and to argue about what's next.
At a glance
What's here.
The shape of this page in six lines.
Don't fine-tune the LLM. Train the agent.
Why domain intelligence belongs in the substrate, not the weights.
AkashaKV at 630K writes/sec
WAL group commit + lock-free buffers — how we 30×'d Redis.
Bi-temporal isn't optional
Why every agent platform needs valid-time + transaction-time.
Working memory ≠ context window
Miller 7±2, decay, binding — what the cognitive science says.
Learning Quarantine
How we sandbox new knowledge against adversarial input.
The Cognitive Cycle, illustrated
Perceive · Attend · Decompose · Reason · Learn · Act.
Reference
Useful links.
referencebash
# Subscribe\n# https://minds.sh/blog.atomSpecs
Details.
At a glance
- Feed
- /blog.atom
- Cadence
- biweekly
- Authors
- engineers, researchers, founders
Guarantees
- Durability
- fsync per WAL commit
- Isolation
- MVCC · snapshot
- Audit
- BLAKE3 Merkle chain
- Encryption
- AES-256-GCM per-namespace
- Concurrency
- 100K+ ops/sec
Get started in an afternoon.
Run anywhere — Akasha Cloud, on-prem, or air-gapped.