The agentic era, distilled.
A curated, source-backed guide to what Databricks announced at DAIS 2026 — organized so you can skim the headlines, dive deep on what matters to your stack, or trace every claim back to official docs.
Event: Data + AI Summit 2026 · Moscone Center, San Francisco · June 15–18, 2026
Compiled: June 17, 2026 (mid-summit) · Sources: official Databricks newsroom, blogs & docs only
This isn't a static report. It's a small repo you run with Claude: ask it questions, and refresh it with new announcements as the summit and its recap posts keep rolling out.
Ask it anything — point Claude at this folder and try:
"Read
dais-2026/00-START-HERE.mdand get oriented." "What changed for our Lakeflow/DLT pipelines?" "Summarize the Genie announcements for a non-technical exec."
Keep it current — just say the magic words ✨
Claude reads _SYNC-RUNBOOK.md, hunts official Databricks sources, dedups
against what's already captured, and appends new findings to UPDATES.md +
feature-status.md — without clobbering your existing files.
| Say this | And Claude will |
|---|---|
| "Sync the Databricks Summit pack" | Find new official content, append to the changelog (safe, additive) |
| …add "full rebuild" | Also fold updates into the main topic files |
| …add "just status" | Only refresh the preview→GA status tracker |
| …add "since June 18" | Only look for content after that date |
Then check UPDATES.md (dated changelog) and feature-status.md
(GA vs Preview at a glance) to see what moved. The runbook is self-contained, so even a brand-new
Claude session can run a sync correctly.
DAIS 2026 wasn't a scattershot product dump. It had a spine:
Agents are now first-class actors on enterprise data — and Databricks rebuilt the platform so they get Choice (any model, format, cloud), Context (governed business meaning), and Control (governance over what agents do, not just what they access).
This pack turns 800+ sessions and a week of announcements into something you can actually use: what shipped, what stage it's in, and what it means — whether you're an existing Spark/MLflow/Genie shop, evaluating Databricks for the first time, or wiring DuckDB and Splink into a governed lakehouse.
No hype. No third-party hot takes as gospel. Every note ties back to documented sources.
| If you want to… | Go to |
|---|---|
| Get oriented in 5 minutes | 00-START-HERE.md |
| See the marquee launches at a glance | 01-headline-launches.md |
| Understand the event themes & keynotes | 00-overview.md |
| Verify links & release stages | 03-sources.md |
| Have some fun — flexes, subtweets & memes 🍿 | the-scoop-zingers-and-memes.md |
Databricks framed the whole summit around governing and building for agents. These three lenses show up in Unity Catalog, Agent Bricks, and Genie alike:
| Theme | What it means | Where to read more |
|---|---|---|
| Control | Govern what agents do, not just what they can access | Unity Catalog & AI Gateway |
| Context | Give agents governed business meaning so they stop guessing | Genie One, Agents & Ontology |
| Choice | No lock-in on model, format, cloud, or region | Open formats & OpenSharing |
dais-2026/
├── README.md ← you are here (overview + how to drive it with Claude)
├── 00-START-HERE.md ← your entry point for the content
├── _SYNC-RUNBOOK.md ← how Claude refreshes the pack + the magic words
├── UPDATES.md ← append-only changelog of every sync
├── feature-status.md ← preview → GA status tracker
├── the-scoop-zingers-and-memes.md ← 🍿 the fun file: flexes, subtweets, meme bait
├── 01-announcements/
│ ├── 00-overview.md ← event facts, keynotes, themes
│ ├── 01-headline-launches.md ← marquee items at a glance
│ ├── 02-unity-catalog-governance.md
│ ├── 03-genie.md ← Genie One / Agents / Ontology
│ ├── 04-lakeflow-spark.md ← DLT → Lakeflow + Spark Declarative Pipelines
│ ├── 05-mlflow.md ← MLflow 3 for GenAI
│ ├── 06-agent-bricks-ai.md ← Agent Bricks, models, Lakebase memory
│ └── 07-open-source-sharing.md ← Iceberg v3, OpenSharing, open formats
├── 02-impact/
│ ├── 01-existing-customers.md ← additive upgrades for current users
│ ├── 02-new-customers.md ← what to evaluate first
│ └── 04-interop-duckdb-splink.md ← off-platform tools + governed tables
└── 03-sources.md ← every link + the dedup ledger for syncs
| Area | Launch | Why it matters |
|---|---|---|
| AI Governance | Unity AI Gateway | One runtime layer to govern every model, MCP, tool, and agent |
| Business AI | Genie One & Genie Agents | Genie Spaces evolve into shareable, autonomous, action-taking agents |
| Context | Genie Ontology | Auto-learned context graph — big accuracy jump on governed semantics |
| Agent dev | Agent Bricks | Full agent platform: any model, governed memory, sandboxes |
| Open formats | Iceberg v3 GA + external write to managed Delta | Deepest open interop; external engines can read and write UC tables |
| Open sharing | OpenSharing → Linux Foundation | Vendor-neutral sharing of AI assets without copy or lock-in |
| Data eng | Spark Declarative Pipelines | DLT becomes an open Apache Spark standard + AI-native authoring |
| ML/GenAI | MLflow 3 for GenAI | Tracing, eval, and observability for agents — even off Databricks |
Full table with release stages (GA / Preview / Beta): headline launches
You're in luck — almost nothing here is disruptive. It's additive and upgrade-shaped. Start with:
- Genie → Genie Agents — your Spaces are the on-ramp
- Lakeflow / Spark Declarative Pipelines — the future of your DLT pipelines
- MLflow 3 for GenAI — tracing & eval for agents
- Impact for existing customers — decisions worth making now
Focus on the platform bet, not the buzzwords:
The open-format story is the bridge:
- Official sources only — databricks.com newsroom, blog, and docs. Third-party links in 03-sources.md are context, not authority.
- 2026 content — where a feature launched earlier and was only re-stated at DAIS 2026, that's flagged in the notes.
- Release stages move fast — GA, Public Preview, Beta, and Private Preview are quoted as Databricks stated them in mid-June 2026. Re-confirm in docs before you build.
| What | Where |
|---|---|
| Keynote livestream & replays | databricks.com/dataaisummit |
| Product announcement blogs | databricks.com/blog |
| Feature status & how-to | docs.databricks.com release notes |
| Session catalog (800+ breakouts) | Summit agenda |
Full annotated link list: 03-sources.md
This is a personal briefing pack, not an official Databricks resource. If you spot a stale release stage or a broken link, open an issue or PR — especially on 03-sources.md.
Built for people who missed the keynotes, sat through them and forgot half of it, or need to explain DAIS 2026 to their team on Monday morning.