Brain Levels & Weekly Digest: Your AI's Progress Report

The invisible growth problem

When your AI brain stores 100 lessons, it's genuinely smarter than when it had 10. It avoids known failure patterns. It recalls the right solution faster. It briefs you with context you'd otherwise have to re-establish manually.

But without a way to see that progress, it's invisible. Developers give up on AI memory systems because they can't measure the value. "Is this actually working?" — the most common question we heard.

Brain Levels and the Weekly Digest make the invisible visible.

Brain Levels: gamifying AI learning

Every Cachly brain has a level. Level 1 starts at 0 lessons. The curve follows a logarithmic progression — early levels are fast (10 lessons per level), later levels slow down (50–100 lessons per level). This mirrors how learning actually works: early wins are easy, deep expertise takes more effort.

The progression reflects how much better your brain's session briefings get compared to a blank slate — measured as the ratio of session-start recall hits to total queries across the last 30 days. It's computed by brain_doctor and displayed in the Brain Visualizer.

The Weekly Digest

Every Monday at 9 AM UTC, Cachly generates a Weekly Brain Digest for each active instance. It's a structured Markdown summary delivered to wherever you prefer: Slack, email, Telegram, or webhook.

## Your AI Brain — Week in Review

**Level:** 34 → 36 (+2 levels)
**New lessons this week:** 23
**Total lessons:** 847

### What your AI learned this week:
- deploy:kubernetes — 8 lessons (most active area)
- fix:env-vars — 4 lessons
- debug:memory-leak — 3 lessons
- api:rate-limiting — 3 lessons

### Top prediction (87% confidence):
Your brain has mastered `deploy:kubernetes`.
Next likely area: `infra:monitoring`
(3 other brains on your team learned this after kubernetes)

### Team knowledge gap:
Your brain does NOT know: `infra:grafana-alerts`
4 teammates' brains know this topic.
Ask your AI: "What does the team know about Grafana alerts?"

### Action recommended:
Run `brain_doctor` to see your full health report.
Your recall hit rate dropped from 71% to 64% —
consider running `memory_crystalize` to consolidate.

The digest is powered by the same GET /api/v1/instances/:id/predictions endpoint that drives the prediction nodes in the Brain Visualizer. The team knowledge gap section comes from the Global Brain API.

Configuring the digest

Set up delivery in the dashboard under Settings → Brain Digest:

• Slack webhook: paste your incoming webhook URL, done
• Email: any address, plain Markdown or HTML
• Telegram: enter your bot token + chat ID
• Webhook: POST to any endpoint, JSON body with the full digest data (not just the Markdown)

Or generate it on-demand via MCP: brain_doctor includes a digest in its output. Run it at the start of any session to get a progress report before you even start working.

Why this matters for teams

The weekly digest is most valuable for engineering managers and team leads. It answers questions that are otherwise impossible to answer:

• What technical areas is my team's AI spending the most time on?
• Are there knowledge silos? (one person's AI knows something nobody else's does)
• Is AI actually being used productively, or is it just answering the same questions repeatedly?
• Which AI assistants have the highest recall hit rates? (proxy for which developers are using the memory system most effectively)

The org-level digest (available on Business and Enterprise) aggregates all instance digests into a single weekly report — the engineering equivalent of a standup, but for your AI workforce.