Videos 9arM9b7JgOo
Develop at Idea Velocity - Jeffrey Lee-Chan, Snapchat
Scene timeline
48 shot(s).
keyframes kept every frame deduplicated
What was stored
- cues
- 193
- whisperx 193
- chunks
- 28
- from 193 cues
- keyframes
- 30
- kept of 48 captured
- frames with text
- 30
- 2,069 lines read
- chapters
- 0
- from the source metadata
- keyframe bytes
- 5.4 MB
- word timings on 193 cues
Provenance
| stage | state | model | started | took |
|---|---|---|---|---|
fetch |
done | — | 2026-08-11 04:09 | 1m 13s |
stt |
done | — | 2026-08-11 04:10 | 16s |
chunk |
done | — | 2026-08-11 04:10 | 0s |
text_embed |
done | — | 2026-08-11 04:10 | 0s |
keyframe |
done | — | 2026-08-11 04:10 | 47s |
ocr |
done | — | 2026-08-11 04:11 | 27s |
frame_embed |
done | — | 2026-08-11 04:11 | 5s |
Frames, and what the machine read
-
- SECTION1.00
- ARCHITECTURE1.00
- The OpenClaw stack and orchestration model1.00
-
- View1.00
- History1.00
- Bookmarks1.00
- Profiles0.99
- Window1.00
- Help1.00
- 90.68
- #0.51
- ⑥0.60
- Q0.89
- 2.0.90
- Sun Mar 29 5:07 PM0.99
- à0.92
- E0.89
- M0.99
- D.0.81
- D.0.52
- D0.67
- featpha0.56
- Safeco0.95
- housh0.99
- WorldAl0.76
- WoridAl0.81
- OPual reo0.69
- Bulding0.98
- C fixdmat0.62
- featavo0.66
- 日0.92
- 目0.96
- 口0.53
- Openclaw Agent Orchestrat:0.97
- smartclaw/docs/opencias0.92
- smartclaw/docs/openclaw0.99
- smartclaw/docs/openclaw0.96
- +0.99
- 3 ao: primary0.87
- C0.57
- docs.google.com / Openclaw Workshop March 29th0.96
- ☆0.80
- 00.59
- 三0.99
- ( Assistant0.97
- Rot's Pad Pho0.83
- $ openclaw: main0.95
- Bookmarks1.00
- 0d0.71
- D m0.69
- D wd0.82
- odev0.95
- Odo0.60
- OPPs0.73
- © odx0.72
- m0.81
- All Bookmarks0.99
- ↓ 0: zbot0.84
- Agent tus complete0.83
- main /.openciow0.72
- feat/z00m-eetinghal0.71
- I3 PR 6442 open0.70
- Claude le walting for you0.94
- Q Menus0.95
- Openclaw Workshop March 29th1.00
- File0.95
- Edit1.00
- View1.00
- Insert0.99
- 90.79
- Format1.00
- 100%1.00
- Tools1.00
- Normal text1.00
- Gemini1.00
- Extensions Help1.00
- Arial1.00
- 10.54
- +1.00
- A0.92
- ○田0.53
- I0.93
- ①0.52
- 回0.86
- 60.95
- Q0.64
- Editing0.98
- Share1.00
- 4 02: setup0.88
- ←1.00
- I' ER.e6.00 open0.67
- Feat/jleectan-0.79
- 402: misc0.93
- dev1774413191* .0.88
- w2: avatar0.93
- Iblog: main0.98
- Activity0.96
- Later1.00
- DMs0.98
- Flles0.95
- G0.53
- 00.84
- Q0.67
- Headings you add to the document wil appear0.99
- here.1.00
- Document tabs1.00
- Tab10.94
- Agenda1.00
- Intro0.99
- Post workshop Q&A if folks interested0.98
- Slides Openclaw:Openclaw Agent Orchestrator extended0.98
- What is Openclaw conceptually0.99
- 15 min intro / demos0.97
- 5 min Q&A1.00
- 20 min workshop setup0.98
- 10 min advanced topic,0.97
- 10 min Q&A1.00
- 90.90
- +0.98
- JAustin0.93
- Ke J0.85
- I2 1R.012 00n0.70
- antig: wc0.94
- feat/wc-b2de0.79
- More1.00
- ...0.70
- Demo1.00
- WorldAl: tos/mvy-site-app-dev-ixf2p72ka-uc a run.app0.71
- Consensus ML:hps//consenus-mlaconveraionld-ior3vk0GDUIGv2iceAc0.75
- 0: primary0.96
- dev2774343982+0.80
- Admin0.99
- Demo: slack + agent orchestrator0.99
- ao: antig0.99
- Video: TODO1.00
- Show PRs merged:1.00
- THIS IS A DEV BUILD0.97
- dav1763892 + /g0.61
- Wi tails0.84
- Prototype: skeptic agent0.99
- Saved items0.98
- Summary PRs: agent orchestrator1.00
- Show omux terminal steering0.99
- Harness skill0.96
- Eloop1.00
- ②0A00.60
- worldsi clsw0.83
- Antigravity - Settings0.98
- 00.97
- JEFFREY LEE-CHAN0.97
- 2026-03-29 17:07:460.98
- OpenClaw · Al Engineer Worlds Fair 20260.96
-
- History1.00
- Help1.00
- d0.58
- 60.97
- Esa0.65
- 2.0.62
- Sun Mar 291.00
- E0.57
- D.0.79
- D0.80
- teatplo0.65
- Safeco0.92
- househ:0.88
- WorldAl0.96
- WorldAl0.96
- O Pul re0.84
- Bulding0.92
- C0.92
- fixdmvat:0.69
- C teat[av0.65
- +0.97
- 日0.97
- 目0.93
- Openclaw Workshop March1.00
- Openclaw Agent Orchestral0.97
- smartclaw/docs/openclaw0.96
- smartclam/docs/openclaw0.96
- smartclaw/doos/openclaw0.98
- +0.99
- 3 ao: primary0.90
- C0.71
- docs google.com / Openclaw Agent Orchestrator + General Al builder topics0.98
- ☆0.85
- 00.63
- 弗0.58
- ( Assistant0.96
- Rot's iPad Pho0.85
- $ openclaw: main0.90
- Bookmarks0.99
- d0.96
- D ws0.82
- D wd0.73
- odev0.89
- 0d0.50
- Osh0.56
- OPRs0.75
- @ odx0.83
- 第c0.61
- m0.82
- All Bookmarks0.99
- Agent turs complene0.90
- File Edit View Insert Format Slide Arrange Tools Extensions Help0.99
- Openclaw Agent Orchestrator + General Al builder topics1.00
- Slideshow1.00
- Share1.00
- ↓ 0: zbot0.82
- Claude is walsing for you0.95
- Q Menus0.87
- +5Fit0.80
- Background1.00
- Layout1.00
- Theme1.00
- Transition1.00
- I1 PR.e642 open0.76
- + 02: setup0.92
- dov1774433191*0.80
- DMs0.93
- I3 28.40.00 open0.68
- a02: misc0.97
- w2: avatar0.87
- Iblog: main0.98
- feat/jleechan0.86
- le is waiting for0.89
- Activity0.92
- Later1.00
- Flles0.96
- G0.58
- 口0.57
- 00.81
- 00.54
- 0000.95
- Openclaw Agent1.00
- Develop at idea velocity1.00
- Orchestrator1.00
- Jeffrey Lee-Chan1.00
- 国0.79
- JAustin0.92
- Ke J0.81
- antig: wc0.95
- More1.00
- ..0.64
- https://www.linkedin.com/in/jeffrey-lee-chan/0.99
- Snapchat Engineering Manager, ex Google0.99
- 口0.67
- I229.8126 000n0.67
- 0: primary0.96
- Admin0.99
- ao: antig0.97
- dov1776463592* ± -/gr0.72
- Claudle is walt0.77
- Enhance this slide X0.96
- Wi tails0.88
- THIS IS A DEV BUILD0.97
- 品0.84
- Click to add speaker notes1.00
- ②0Δ00.58
- Antigravity - Settings0.99
- 00.94
- JEFFREY LEE-CHAN0.99
- 2026-03-29 17:08:310.98
- OpenClaw · Al Engineer Worlds Fair 20260.95
-
- docs.google.com - To exit full screen, press0.99
- Openclaw Agent1.00
- Orchestrator1.00
- Develop at idea velocity1.00
- Jeffrey Lee-Chan0.98
- Snapchat Engineering Manager, ex Google1.00
- https://www.linkedin.com/in/jeffrey-lee-chan/1.00
- JEFFREYLEE-CHAN1.00
- 2026-03-2917:08:370.98
- OpenClaw · Al Engineer Worlds Fair 20260.98
-
- HowIt Works0.94
- A single natural-language message in Slack triggers the full autonomous pipeline — from intent to green checks.0.99
- Agent1.00
- Describe1.00
- Context+0.97
- Cl& Review0.96
- Proof-First1.00
- Worktrees1.00
- Memory1.00
- Each stage is fully automated and evidence-backed — no manual re-entry of context, no babysitting Cl, no chasing reviewers.0.99
- JEFFREYLEE-CHAN1.00
- 2028-03-2917:09:030.99
- OpenClaw· Al Engineer Worlds Fair 20260.98
-
- SECTION1.00
- LIVE DEMO1.00
- Pipeline in action — Slack intent to green checks0.99
-
- Demo (Video)0.99
- Rotrs Pad Pho0.84
- A single natural-language message in Slack triggers the full autonomous pipeline — from intent to1.00
- green checks.1.00
- Openclaw slack coding1.00
- 白0.58
- Activity1.00
- # Thread =0.87
- Top 3 1d prioritieas mportant + urrntlyeno open Pk tled to the task lbused on ddocs.0.84
- and open P1 beads0.97
- Why top: explicitly marked as Next PRr° and High risk in roadmap: reluces core ddive0.88
- rev-10c4 - unify streaming/non-streaming LLM path0.96
- 14 nim0.71
- PR status: no open PR cume0.99
- Evidence bead rev-ba P1, pent " ses load balancer IP..fa with X-Forwarded-For + ProxyyFix0.79
- revwg2r - agent selection ircsit breaker Infinite lop prevention)0.85
- Durity impacting isse (global bucket / DoS rik va LB IP handing)g0.89
- rate limiter IP spofing vulnerablity0.90
- ently for this bead ID.1.00
- ontly fo this bead ID.0.84
- my safety issue: prevents repeated same-agent loop failures.0.96
- pnd + bead rev-lac4 (P1. oponl.0.80
- •PR status: no opem PR cumently for this bead ID.0.93
- If you want, mext I can hand these 3 to AO as separate tracks (sorlaorchitect ) and post back0.95
- appear.0.99
- @openciaw make agento0.85
- 0:42/4:011.00
- a @0.67
- JEFFREYLEE-CHAN1.00
- 2026-03-2917:11:000.98
- OpenClaw · Al Engineer Worlds Fair 20260.96
-
- docs.google.com - To exit full screen, press0.99
- Demo(Video)1.00
- RobsiPadPto0.86
- A single natural-language message in Slack triggers the full autonomous pipeline — from intent to1.00
- green checks.0.99
- Openclaw slack coding1.00
- Activity0.92
- JEFFREYLEE-CHAN1.00
- 2026-03-2917:11:401.00
- OpenClaw· Al Engineer Worlds Fair 20260.98
-
- What is Openclaw?0.99
- OpenClaw - Autonomous Al Coding Architecture1.00
- Rotis Pad Pho0.84
- Central Al0.98
- Data Sources & Services0.99
- Data Sources & Services0.98
- Beyond any particular code or repo the0.99
- Error Logs0.99
- Google Docs1.00
- concept is important0.98
- Customer Complaints0.99
- Google Calendar0.99
- Me1.00
- OpenClaw0.94
- Slack1.00
- Frictionless communication1.00
- Meeting Notes1.00
- Google Calendar1.00
- Memory1.00
- Emails1.00
- Slack, WhatsApp, Telegram,0.99
- Otition0.94
- Database1.00
- Access to central data sources and1.00
- Skills1.00
- Cmux Terminals0.92
- Slack1.00
- services1.00
- Orchestrator Agents1.00
- Claude Code1.00
- The more access you give to everything you1.00
- Agent0.91
- Codex0.99
- Codex0.99
- can access the more powerful ie. Docs,1.00
- Gerrini0.93
- Gmail, Slack1.00
- Codex0.96
- Memory1.00
- Sub-agents & eif-Managing Agents0.96
- It remembers what your goals/desires/past1.00
- actions are ideally replacing more and more0.98
- of your decisions1.00
- JEFFREY LEE-CHAN0.97
- 2026-03-29 17:12:020.98
- OpenClaw · Al Engineer Worlds Fair 20260.96
-
- SECTION1.00
- CODE WALKTHROUGH1.00
- Inside the multi-agent loop1.00
-
- Agent stack1.00
- Jeff stack / open source0.98
- OpenClaw1.00
- 1. Openclaw0.99
- 2. Agent managers1.00
- Cmux Terminals1.00
- + Orchestrator0.99
- a. CMUX terminals OR1.00
- Agents1.00
- b. Agent Orchestrator managers0.99
- 3. Agent Orchestrator workers (claude code, codex,0.99
- cursor cli)1.00
- Agents1.00
- Claude code / codex / etc0.96
- 1. Managed agents (agents managed by above aka1.00
- claude teams feature)0.99
- 2. Subagents0.98
- Codex0.97
- JEFFREYLEE-CHAN1.00
- 2026-03-29 17:15:470.99
- OpenClaw · Al Engineer Worlds Fair 20260.97
-
- Al builder evolution1.00
- Rot's iPad Pho0.88
- THE EVOLUTION OF AI IN CODING: FROM TOOLS TO SELF-EVOLUTION0.99
- STAGE A - FOUNDATIONS0.97
- STAGE B- AGENTIC SHIFT0.96
- STAGE C - ADVANCED0.98
- STAGE D – SELF-EVOLUTION0.98
- & IDE AUGMENTATION0.98
- & AUTONOMY0.99
- ORCHESTRATION & SELF-EVOLUTION0.99
- (no one yet)1.00
- No Al0.96
- Claude - terminal and review code0.98
- Harness engineering - maximize0.99
- Self evolving systems that1.00
- - no more IDE0.96
- agent autonomy and minimize0.98
- automatically improve their code0.97
- human interaction1.00
- MAX AUTONOMY0.97
- Auto complete - copilot in IDE0.98
- Coding parallel - usually0.97
- worktrees and multi agent1.00
- Orchestrators - manage 10-200.98
- Openclaw - sort of0.96
- agents max parallelize1.00
- Cursor - code assist in IDE0.96
- orthogonal but1.00
- big multiplier1.00
- Cursor1.00
- 10-201.00
- AUTO-IMPROVE1.00
- Looping agents1.00
- have them run0.99
- for 3 hr+ of0.99
- planned work1.00
- 3h+1.00
- NO ONE YET1.00
- JEFFREYLEE-CHAN1.00
- INFOGRAPHIC DESIGN1.00
- 2026-03-2917:16:201.00
- OpenClaw · Al Engineer Worlds Fair 20260.96
-
- Agent stack1.00
- Jeff stack / open source0.97
- OpenClaw1.00
- 1. Openclaw0.99
- 2. Agent managers1.00
- Cmux Terminals1.00
- + Orchestrator0.99
- a. CMUX terminals OR1.00
- Agents1.00
- b. Agent Orchestrator managers0.99
- 3. Agent Orchestrator workers (claude code, codex,1.00
- cursor cli)1.00
- Agents1.00
- Claude code / codex / etc0.94
- 1. Managed agents (agents managed by above aka0.99
- claude teams feature)0.98
- 2. Subagents0.97
- JEFFREY LEE-CHAN0.98
- 2026-03-2917:17:081.00
- OpenClaw · Al Engineer Worlds Fair 20260.97
-
- SECTION1.00
- PRACTICAL OUTCOMES0.99
- WorldAI RPG and Consensus ML — production lessons0.99
-
- View1.00
- History1.00
- Profiles0.98
- Tab1.00
- Window1.00
- Help1.00
- #0.63
- ⑥0.53
- Sun Mar 29 5:28 PM0.99
- à0.70
- E0.85
- D.0.74
- D0.69
- D0.62
- teatpka0.62
- Safeco0.98
- WorldAl0.93
- WorldAl0.80
- O Pul re0.86
- Bulding0.94
- C fixdmvat:0.65
- C feat(avo0.71
- V0.60
- 日0.94
- Openclaw Workshop March0.99
- Openclaw Agent Orchestrat:0.95
- claude-commands/.claude/0.98
- smartclam/docs/openclaw0.98
- smartclaw/docs/openclaw0.99
- ao: primary0.88
- github.com / claude-0.98
- ☆0.98
- 00.52
- B0.54
- ( Assistant0.96
- $ openclaw: main0.95
- Bookmarks1.00
- 0d0.69
- D wd0.90
- odev0.87
- dp1.00
- OPRs0.76
- @ odx0.83
- 第c0.69
- All Bookmarks0.95
- main / openciow0.73
- Agent turn complene0.87
- Files1.00
- 4layer.md0.98
- ↑ Top0.84
- 4 0: zbot0.90
- P main0.92
- Preview1.00
- Code1.00
- Blame1.00
- 115 tines (85 lac) - 3.89 K80.90
- C Raw0.73
- I2 E8 6442 open0.64
- 4 02: setup0.86
- dev1774433191* .0.83
- w2: avatar0.90
- Home0.97
- DMs0.99
- claude0.94
- Q. Go to file0.88
- I claude-plugin0.90
- agents0.94
- 1. Unit Tests ( $PROJEC_A0OT/tests/)0.94
- Minimal Repro Ladder0.99
- Execute tests in this order, stopping only when blocker is conclusively reproduced:0.99
- 3 88.40.00 open0.67
- feat/jleechan0.91
- a02: misc0.96
- I1 28.8126 00n0.72
- Iblog: main0.98
- antig: wc0.93
- 0: primary0.96
- dev2774343982± .0.78
- feat/wc-b20e0.81
- main- -/aeojecta0.62
- Activiny0.87
- More0.99
- Admin0.99
- Later1.00
- Flles1.00
- ..0.65
- G0.68
- Q0.64
- commands0.96
- _copilot_modules0.94
- spec-kit0.91
- I _shared0.83
- cerebras0.90
- I tests0.95
- .ruff.toml0.92
- 4layer.md1.00
- CLAUDE.md0.99
- ENHANCED_ALIASES.md0.99
- LEARN_ENHANCEMENT_SUM...0.99
- MEMORY_INTEGRATION.md0.99
- 2. End-to-End Tests (SPROJECT_RO0T/tests/test_end2end/)0.97
- 3. MCP/HTTP API Tests (real local server, testing_mcp/)0.97
- 4. Browser Tests (final escalation, testing_ui/)0.98
- ./vpython -m pytest sPROJECT_R0OT/tests/test_[relevant).py -q0.95
- ./vpython -# pytest sPROJECT_R00T/tests/test_end2end/test_[featurel_end2end.py -q0.97
- ./vpython testing_mcp/[domain]/test_[feature]_real.py0.98
- ./vpython testing_ui/[domain]/test_[feature]_browser.py0.99
- O0.57
- 00.52
- 00.51
- THIS IS A DEV BUILD0.98
- dav17643892 + -/p0.59
- W: talls0.82
- ao: antig0.98
- README.md0.99
- README_EXPORT_TEMPLATE...0.99
- SLASH_COMMAND_RECOGNIT...0.99
- _copilot_reference.md0.92
- adde2e.md1.00
- agento_report.md0.97
- Action Steps1.00
- 2. Start at Layer 1 - Run unit tests first0.99
- 5. Report results - Document which layer reproduced the issue0.99
- 3. Climb ladder - Progress to next layer only if current layer passes1.00
- 1. Identify blocker - Determine which feature/component is failing0.99
- 4. Collect evidence - Record full absolute paths to evidence bundles0.99
- ②0Δ00.58
- orldai claw0.83
- Antigravity - Settings0.97
- 00.97
- JEFFREY LEE-CHAN0.98
- 2026-03-2917:28:411.00
- OpenClaw · Al Engineer Worlds Fair 20260.97
-
- View1.00
- History1.00
The page's on-screen-text budget of 600
lines is spent, so the last cards in this grid list fewer lines than they
hold. Narrow the page with ?frames= to read them.
Transcript
193 cues· 2,450 words· 12,472 chars
- 0:01 workshop setup.
- 0:02 So there are some people who have never installed OpenClaw, and I want to get them set up.
- 0:09 And then the other people want to try an experimental thing, which I haven't even fully gotten working.
- 0:15 But I think for more advanced people, that'll be cool to get a sort of staging environment and run to OpenClaw.
- 0:21 So I think if I'm not too busy helping new people set up, then I'll try to help people finish debugging the last part.
- 0:29 Or maybe someone else will get it.
- 0:32 And I think 10 minutes Q&A, 10 minutes some more advanced topics, and then just more Q&A.
- 0:39 And probably after every five minutes or so, we can do a few minutes of discussion.
- 0:44 So I'll just set some timers to make sure I don't go off track.
- 0:49 All right, let me start.
- 1:02 Can everyone see this presentation?
- 1:07 Yes.
- 1:08 All right, cool.
- 1:09 Yep.
- 1:11 So this kind of describes my setup.
- 1:16 I'll just keep it really high level for this part, where when everything works, maybe like 70% of the time, you describe your task.
- 1:25 And then what's nice is OpenClaw will have context and memory about what you're talking about.
- 1:31 So for example, let's say I'm like, all right, fix this thing.
- 1:35 It's a very brief Slack message, but OpenClaw can remember what I've asked it to do before.
- 1:41 So for example, I'm working on this thing I'm calling a skeptic agent, and it's a custom code review thing.
- 1:47 But if I'm like, all right, fix the skeptic agent or whatever, it kind of knows what that is, which is nice.
- 1:52 You don't have to re-explain a lot.
- 1:55 I use multiple agents with work trees.
- 1:57 That's very key for parallelization.
- 2:06 integration.
- 2:07 But you can kind of see, you can talk to it and be pretty brief and then, you know, usually get reasonable outcomes, but sometimes you'll have to debug it.
- 2:16 So then it's like, all right, there's some timeout thing.
- 2:19 I'm like, give it a longer timeout, right?
- 2:24 And then I asked it here, like, you know, what are the priorities and what we're working on?
- 2:28 And I'm reading it here.
- 2:31 And then I might have just said, like,
- 2:35 I just make code for it.
- 2:38 Another interesting thing about this is you can see from the types of responses I give, they're not that special as in I actually think an agent could replace me.
- 2:51 That's what I'm working on right now to get things even more autonomous.
- 2:57 You'll see this with LLMs, you're working with them and they'll be like, hey, I've got this thing ready and I'm like, okay, we'll run the test.
- 3:06 all right so i'll show the next slide now oops okay we'll go into some concepts right now like what is open claw so i think what's cool about open claw is it's not just about a particular repo or the code but more of the concept so you know i've got this cool super diagram but
- 3:33 First concept is frictionless communication.
- 3:36 I know we have other things like cloud co-work or whatever, but I think they still have a similar concept where it's like, can you easily talk to your AI versus you've got a remote desktop into your computer or you have to go sit down.
- 3:51 That enables a certain really high level of velocity.
- 3:56 Then I think the central axis is very important.
- 4:00 You know, there's a spectrum.
- 4:02 Like, I kind of go like a little more cowboy and I give it action.
- 4:12 OpenClaw has these agent orchestrator managers that manage your workers for you.
- 4:17 Or sometimes when I want more personal control, I use CMUX terminals.
- 4:21 So CMUX is like a terminal program, but it's pretty good for parallelization and AI development.
- 4:27 Then I used the open source framework that I forked, agent orchestrator to do the workers.
- 4:33 And then once it gets to here, this part is not really controlled by me as much anymore.
- 4:38 So I'm calling them managed agents.
- 4:41 But basically, you have a worker that runs Cloud Code.
- 4:45 And Cloud Code itself can run agents.
loading