vidtheque.

read-only demo

Videos jtzh-GBXBWc

The Factory That Dreams: 39 AI Agents, No Framework - Rushabh Doshi, Machinecraft

index_state ready data_status ok

AI Engineer· published 2026-07-11· 0:09:58· en· indexed 2026-08-11 04:02

Open on YouTube

Scene timeline

  1. Shot 0, 0:00 to 0:28, 1 of 1 keyframes kept
  2. Shot 1, 0:28 to 0:55, 1 of 1 keyframes kept
  3. Shot 2, 0:55 to 1:25, 1 of 1 keyframes kept
  4. Shot 3, 1:25 to 1:49, 1 of 1 keyframes kept
  5. Shot 4, 1:49 to 2:23, 1 of 1 keyframes kept
  6. Shot 5, 2:23 to 2:41, 1 of 1 keyframes kept
  7. Shot 6, 2:41 to 3:16, 1 of 1 keyframes kept
  8. Shot 7, 3:16 to 3:51, 1 of 1 keyframes kept
  9. Shot 8, 3:51 to 4:12, 1 of 1 keyframes kept
  10. Shot 9, 4:12 to 4:42, 1 of 1 keyframes kept
  11. Shot 10, 4:42 to 4:59, 1 of 1 keyframes kept
  12. Shot 11, 4:59 to 5:11, 1 of 1 keyframes kept
  13. Shot 12, 5:11 to 5:43, 1 of 1 keyframes kept
  14. Shot 13, 5:43 to 6:02, 1 of 1 keyframes kept
  15. Shot 14, 6:02 to 6:24, 1 of 1 keyframes kept
  16. Shot 15, 6:24 to 6:37, 1 of 1 keyframes kept
  17. Shot 16, 6:37 to 6:50, 1 of 1 keyframes kept
  18. Shot 17, 6:50 to 7:24, 1 of 1 keyframes kept
  19. Shot 18, 7:24 to 7:31, 1 of 1 keyframes kept
  20. Shot 19, 7:31 to 7:59, 1 of 1 keyframes kept
  21. Shot 20, 7:59 to 8:18, 1 of 1 keyframes kept
  22. Shot 21, 8:18 to 8:43, 1 of 1 keyframes kept
  23. Shot 22, 8:43 to 9:08, 1 of 1 keyframes kept
  24. Shot 23, 9:08 to 9:35, 1 of 1 keyframes kept
  25. Shot 24, 9:35 to 9:57, 1 of 1 keyframes kept

25 shot(s).

keyframes kept every frame deduplicated

What was stored

cues
132
whisperx 132
chunks
19
from 132 cues
keyframes
25
kept of 25 captured
frames with text
25
110 lines read
chapters
0
from the source metadata
keyframe bytes
3.8 MB
word timings on 132 cues

Provenance

Each pipeline stage, its state and the model that produced it
stage state model started took
fetch done — 2026-08-11 03:59 1m 23s
stt done — 2026-08-11 04:01 12s
chunk done — 2026-08-11 04:01 0s
text_embed done — 2026-08-11 04:01 1s
keyframe done — 2026-08-11 04:01 30s
ocr done — 2026-08-11 04:02 7s
frame_embed done — 2026-08-11 04:02 4s

Frames, and what the machine read

  • 0:11 #0 done8 line(s)

    shot 0·sharpness 1417.0

    1. IRA0.99
    2. MACHINECRAFT1.00
    3. IRA-FORKMY0.95
    4. BRAIN1.00
    5. How a 100-person factory taught itself to remember - and built a 36-agent AI that0.99
    6. runs its go-to-market.1.00
    7. Rushabh Doshi0.99
    8. MachinecraftAI Engineer World's Fair1.00

  • 0:36 #1 done4 line(s)

    shot 1·sharpness 1368.0

    1. 011.00
    2. THE SETUP0.95
    3. WHAT A FACTORY ACTUALLY I!0.96
    4. The company isn't the machines. It's the knowledge.0.99

  • 1:13 #2 done3 line(s)

    shot 2·sharpness 2467.1

    1. 02· THE PROBLEM0.98
    2. TWO BRAINS, ONE REVOLVING DOOR1.00
    3. Every time someone left, the company forgot a little more.1.00

  • 1:44 #3 done6 line(s)

    shot 3·sharpness 2147.1

    1. 031.00
    2. THE1.00
    3. IDEA1.00
    4. DON'T WRITE IT DOWN — GROW0.99
    5. BRAIN1.00
    6. Not a chatbot. A twin of the company.0.99

  • 1:56 #4 done3 line(s)

    shot 4·sharpness 2304.7

    1. 04 · THE STAKES0.95
    2. ONE MACHINE, SEVEN WORLDS0.99
    3. Trays · spas · toilets · EV panels · medical · packaging0.96

  • 2:28 #5 done3 line(s)

    shot 5·sharpness 2097.0

    1. 05·STEP ONE0.96
    2. FEED IT EVERYTHING0.99
    3. Years of quotes, drawings, schedules, emails. Our internet, not the internet.0.99

  • 3:09 #6 done3 line(s)

    shot 6·sharpness 2461.9

    1. 06· THE PLOT TWIST0.96
    2. WE NEVER TRAINED A MODEL0.98
    3. Chunk it → vectors in Qdrant → relationships in Neo4j → our own retrieval1.00

  • 3:27 #7 done4 line(s)

    shot 7·sharpness 2174.1

    1. 07·ARCHITECTURE1.00
    2. WE GAVE IT A BODY1.00
    3. Senses· a gut· memory0.98
    4. · a dream cycle · an immune system0.96

  • 4:05 #8 done4 line(s)

    shot 8·sharpness 1824.8

    1. 081.00
    2. THE TEAM0.97
    3. WHY NOT ONE GENIUS?0.99
    4. One prompt that does everything does everything badly.1.00

  • 4:36 #9 done3 line(s)

    shot 9·sharpness 4574.1

    1. MEET THE CAST0.96
    2. THIRTY-SIX AGENTS, ONE JOB EACH0.99
    3. Athena · Prometheus · Argus · Clio· Calliope · Plutus · Hephaestus · Vera · Mnemon0.95

  • 4:49 #10 done4 line(s)

    shot 10·sharpness 2127.1

    1. 101.00
    2. HOW THEY DECIDE0.99
    3. THEY ARGUE IT OUT0.98
    4. A boardroom that never sleeps and has no ego.0.99

  • 5:10 #11 done3 line(s)

    shot 11·sharpness 3598.8

    1. 11·THE JOB0.96
    2. ONE BRAIN RUNS GO-TO-MARK0.99
    3. Everything between 'a stranger exists'and 'they're a customer.0.99

  • 5:15 #12 done4 line(s)

    shot 12·sharpness 4750.3

    1. 12 · IN PRACTICE0.94
    2. NINE REAL JOBS, EVERY DAY0.99
    3. Outbound·briefs1.00
    4. quotes · swipe-to-send · revival · inbound · CRM · research0.97

  • 5:45 #13 done3 line(s)

    shot 13·sharpness 2653.9

    1. 13·THE COCKPIT0.99
    2. ONE CURSOR TAB0.99
    3. You ask; it reaches out with a dozen hands - and shows you before sending.0.99

  • 6:11 #14 done6 line(s)

    shot 14·sharpness 4446.6

    1. 141.00
    2. THE STACK0.99
    3. A REAL STACK, NOT A DEMO0.98
    4. Vector + graph + CRM DBs · 3 model providers · ingest · comms0.96
    5. observability1.00
    6. #0.67

  • 6:31 #15 done3 line(s)

    shot 15·sharpness 2085.0

    1. 15 · THE SURFACE0.98
    2. 213 TOOLS, ONE RULE1.00
    3. Ira drafts. Humans send.1.00

  • 6:42 #16 done3 line(s)

    shot 16·sharpness 2017.4

    1. THE GOLDFISH PROBLEM0.99
    2. MOST AI FORGETS YOU0.99
    3. Brilliant for thirty seconds. Then the tab closes.1.00

  • 7:20 #17 done11 line(s)

    shot 17·sharpness 4247.8

    1. 171.00
    2. ENGINEERED1.00
    3. MEMORY1.00
    4. MEMORY, IN LAYERS0.97
    5. Working · pinned facts · semantic · episodic · relationship0.97
    6. procedural1.00
    7. MW0.73
    8. who they are0.96
    9. relationship warmtk0.95
    10. open commitments0.98
    11. 0001.00

  • 7:28 #18 done3 line(s)

    shot 18·sharpness 1869.6

    1. OVERNIGHT1.00
    2. AND THEN IT DREAMS1.00
    3. The thing literally gets smarter while you sleep.1.00

  • 7:43 #19 done9 line(s)

    shot 19·sharpness 4066.2

    1. 191.00
    2. THE NIGHT0.98
    3. SHIFT1.00
    4. WHAT IT DOES AT NIGHT0.99
    5. Replay1.00
    6. consolidate1.00
    7. prune1.00
    8. · learn skills · write the dream report · wake up0.98
    9. 000.56

  • 8:05 #20 done4 line(s)

    shot 20·sharpness 2010.3

    1. 201.00
    2. THE CONSCIENCE1.00
    3. A CONSCIENCE, NOT A FILTER0.99
    4. Written from the principles of a Jain family business.1.00

  • 8:38 #21 done4 line(s)

    shot 21·sharpness 2415.1

    1. 21·THE GUARDRAILS0.99
    2. FIVE OLD IDEAS, RUNNING IN PROD1.00
    3. Many-sidedness·0.99
    4. qualify claims · do your duty· truth over optics· serve each other0.97

  • 8:58 #22 done6 line(s)

    shot 22·sharpness 2563.8

    1. 22· THE MONEY PART0.98
    2. CHEAPER THAN A NICE WATCH0.97
    3. No training bill · $320K agency quote → ~$30K · ~$2K/month to run0.98
    4. BUILD1.00
    5. INGESTION0.96
    6. FUEL1.00

  • 9:17 #23 done3 line(s)

    shot 23·sharpness 2423.9

    1. THE MOVE0.99
    2. FORK AN EMPTY BRAIN0.97
    3. Brain OS ships blank. You pour your own truth in.1.00

Transcript

132 cues· 1,394 words· 7,805 chars

  1. 0:01 Okay.
  2. 0:01 I want to tell you a story about a factory that taught itself how to remember.
  3. 0:07 Hi, I'm Rushabh.
  4. 0:08 I run Machinecraft, a hundred people factory in India.
  5. 0:12 No data science team, no ML budget, none of that.
  6. 0:15 And somehow we ended up building a 36 AI agent that runs our entire go-to market.
  7. 0:21 I think that's still a little ridiculous.
  8. 0:25 Let me show you how it happened and why you can do the same thing.
  9. 0:30 So here's the thing about our company.
  10. 0:32 From the outside, it looks like machines and metal.
  11. 0:35 But the actual company, the part that matters is in the machines, is the knowledge.
  12. 0:40 Who the customer is, what we quoted them in 2019, why that one machine needed that beard custom tweak.
  13. 0:47 And for three generations, all of that lived in exactly three brains.
  14. 0:51 Initially, my grandfather's, then my father's, and now mine.
  15. 0:57 which is a genuinely terrifying way to run a company when you sit with it.
  16. 1:02 A lot of people have joined us.
  17. 1:04 People have left us.
  18. 1:05 The revolving door never stopped.
  19. 1:08 And every single time someone walked out, a chunk of our brain walked out with them.
  20. 1:15 We weren't scared of the competitors.
  21. 1:17 We were scared of forgetting or waking up one day and realizing the whole company only existed inside two increasingly tired heads.
  22. 1:27 so i had an idea i'll be honest it sounded insane first but what if instead of writing the knowledge down in some document nobody ever reads what if we grew a brain that just held it not a chatbot you poke at a twin of the company i didn't hire a sales team i tried to build one a quick detour because you need to know how messy this is
  23. 1:55 We make thermoforming machines.
  24. 1:57 They heat up a plastic sheet and shape it.
  25. 2:01 Same core machine, but it ends up making hydroponic form trays, spa bathtubs, EV car panels, medical casings, and even packaging.
  26. 2:10 Seven totally different worlds, seven totally different buyers.
  27. 2:15 So this brain couldn't just memorize a brochure.
  28. 2:18 It had to know which universe a given customer lives in.
  29. 2:24 Step one was almost boringly simple.
  30. 2:27 Feed it everything, and I mean everything.
  31. 2:30 Years of quotes, drawings, payment schedules, timelines, email threads, hundreds of gigabytes of our own private history.
  32. 2:38 Not the public internet, our internet.
  33. 2:42 And here's the plot twist, the part that surprises every engineer I tell this to.
  34. 2:48 We never trained a model.
  35. 2:50 No GPUs humming in the basement, no fine tuning.
  36. 2:53 We just looked at all the history, chopped it into bite-sized chunks, and let offshore models read it and pull out the facts.
  37. 3:03 We stored the meaning of each chunk as vectors and relationships.
  38. 3:07 Who's connected to what as a graph?
  39. 3:10 The brain is in a smarter model.
  40. 3:12 It's actually a really, really well-organized memory.
  41. 3:17 Now this is where it gets a little weird in a good way.
  42. 3:22 We stopped thinking of era as a software and started thinking of it as something we were raising.
  43. 3:27 So we gave it a body modeled on biology, senses to figure out who it's talking to, a gut to digest the documents into facts, a memory, a dream cycle, an immune system to fight off bad information.
  44. 3:42 Why biology?
  45. 3:43 Well, because evolution already spent a billion years solving, how do you stay coherent over time?
  46. 3:50 We just copied the homework.
  47. 3:52 Okay so the big question, why 36 agents instead of one genius mega prompt?
  48. 3:58 Because, and you already know this if you've ever tried it, one prompt that's supposed to do everything ends up doing everything badly.
  49. 4:07 So ERA isn't one mind, it's a pantheon, a whole cast of specialists.
  50. 4:13 Each one has exactly one job.

Open at this second