1
00:00:00,900 --> 00:00:03,729
57 files on one research question:

2
00:00:04,100 --> 00:00:07,871
papers, spreadsheets,
Word tables and slide decks.

3
00:00:08,973 --> 00:00:11,479
More files than a browser
chat takes at once.

4
00:00:11,991 --> 00:00:17,304
Here, a team of AI agents reads them all,
and checks every number against its page.

5
00:00:19,600 --> 00:00:24,579
AI for Research Efficiency.
Episode 3: Agents at work.

6
00:00:26,700 --> 00:00:30,389
Last time, you installed an agent
and asked it a first question.

7
00:00:31,094 --> 00:00:34,837
In this episode, a literature
review by several agents at once.

8
00:00:35,287 --> 00:00:39,590
You write the rules, give each agent
one job, and check what they find.

9
00:00:41,300 --> 00:00:45,892
The question comes from Arkansas, which
plants nearly half of the country's rice.

10
00:00:46,594 --> 00:00:51,099
Rice fields are usually kept flooded,
and flooded soil gives off methane.

11
00:00:52,012 --> 00:00:55,932
Alternate wetting and drying lets
the water drain away between floods.

12
00:00:57,135 --> 00:01:00,903
How much methane and water does
that save? What happens to the yield?

13
00:01:01,234 --> 00:01:04,621
And do the Arkansas results agree
with the rest of the world?

14
00:01:07,700 --> 00:01:12,667
The folder holds copies of 57 files
that are free to share, in 8 formats.

15
00:01:13,038 --> 00:01:16,181
The PDFs alone run to 500 pages.

16
00:01:17,523 --> 00:01:20,590
A browser chat takes 10
to 40 files at a time,

17
00:01:20,721 --> 00:01:22,735
and it can't reach the
folder on your computer.

18
00:01:23,146 --> 00:01:26,594
And this job needs every
number traced back to its page.

19
00:01:29,300 --> 00:01:30,584
First, the rules.

20
00:01:31,036 --> 00:01:33,172
They go in a plain text
file in the folder,

21
00:01:33,263 --> 00:01:35,509
which the agent reads
before it starts.

22
00:01:36,490 --> 00:01:39,473
Claude Code reads a
file called CLAUDE.md.

23
00:01:40,245 --> 00:01:43,488
Codex and Antigravity read AGENTS.md.

24
00:01:44,761 --> 00:01:46,947
Write it as you'd brief
a research assistant:

25
00:01:47,198 --> 00:01:50,446
the question, what to extract,
and how to cite it.

26
00:01:51,367 --> 00:01:55,271
Every number with its page,
sheet and cell, or slide.

27
00:01:55,602 --> 00:01:59,456
Quotes, word for word.
Anything uncertain, marked.

28
00:01:59,786 --> 00:02:01,948
And never change the source files.

29
00:02:05,300 --> 00:02:06,463
Then the jobs.

30
00:02:06,874 --> 00:02:09,390
Each kind of agent gets a
short description of its own:

31
00:02:09,741 --> 00:02:12,127
a reader, a checker and a writer.

32
00:02:13,116 --> 00:02:14,698
Readers pull out the evidence.

33
00:02:15,089 --> 00:02:18,324
Checkers trust nothing:
they open every source again.

34
00:02:18,715 --> 00:02:21,739
The writer drafts the answer
from checked rows only.

35
00:02:24,500 --> 00:02:27,429
Then one request, in plain words.

36
00:02:31,232 --> 00:02:33,735
The run is long,
so it works in auto mode,

37
00:02:33,825 --> 00:02:38,370
the faster speed from episode two.
That's why the folder holds only copies.

38
00:02:41,300 --> 00:02:44,068
The lead agent splits
the files into 8 batches,

39
00:02:44,159 --> 00:02:47,066
by topic,
and starts 8 readers at once.

40
00:02:48,387 --> 00:02:50,351
Each reader works
through its own batch.

41
00:02:50,822 --> 00:02:55,040
From a PDF, it takes the text
with the table columns in place,

42
00:02:55,090 --> 00:02:57,846
or looks at the page as an image.

43
00:02:57,896 --> 00:03:00,941
For Word, Excel and PowerPoint,
it writes a short Python script.

44
00:03:02,102 --> 00:03:06,073
Every finding becomes a row:
one number, for one outcome,

45
00:03:06,204 --> 00:03:10,073
with its file, its exact place,
and the words it came from.

46
00:03:11,414 --> 00:03:17,377
In under 6 minutes, the readers are
done: 667 rows, from 40 of the files.

47
00:03:17,527 --> 00:03:20,098
The other 17 had
nothing on the question,

48
00:03:20,249 --> 00:03:23,090
or only pictures,
and the readers said which.

49
00:03:26,900 --> 00:03:29,461
Then 8 checkers, again at once.

50
00:03:30,032 --> 00:03:35,275
Each one opens every cited page, cell
or slide, and compares it with the row.

51
00:03:36,665 --> 00:03:41,908
573 rows are confirmed,
94 corrected, and none rejected.

52
00:03:42,438 --> 00:03:47,040
Most corrections are small: the right
page, or how a study is described.

53
00:03:48,420 --> 00:03:49,742
One changes the meaning.

54
00:03:50,313 --> 00:03:55,249
A reader filed an EPA sentence, "as
much as 50 percent", as a comparison:

55
00:03:55,780 --> 00:04:00,626
flooding fields in winter, or not.
And it said the study was in California.

56
00:04:02,090 --> 00:04:03,691
The checker read the page again.

57
00:04:04,201 --> 00:04:08,214
The sentence gives the share of a year's
methane released during winter flooding,

58
00:04:08,364 --> 00:04:12,907
and the page never says where. The
right number, but the wrong claim.

59
00:04:17,300 --> 00:04:20,407
Last, the writer drafts the
answer from checked rows.

60
00:04:20,838 --> 00:04:24,376
The lead agent then checks that
every statement carries its citation,

61
00:04:24,547 --> 00:04:26,250
and fixes the ones that don't.

62
00:04:27,417 --> 00:04:32,844
Across 10 meta-analyses, drying
cut methane by 31 to 62 percent:

63
00:04:33,155 --> 00:04:34,837
about half, on average.

64
00:04:35,777 --> 00:04:39,527
It saved a quarter to two fifths of
the water, depending on what's counted.

65
00:04:39,998 --> 00:04:43,346
Yield held on average,
unless the drying was severe.

66
00:04:44,269 --> 00:04:48,742
Nitrous oxide went up, so the climate
gain is smaller than the methane cut.

67
00:04:49,883 --> 00:04:55,007
And in Arkansas, field trials saved 34
to 50 percent of the irrigation water,

68
00:04:55,057 --> 00:04:59,411
with no loss of yield: in line with the
world's studies of irrigation alone.

69
00:04:59,761 --> 00:05:01,492
But the folder has no Arkansas

70
00:05:01,542 --> 00:05:04,364
measurement of nitrous oxide
under this practice.

71
00:05:07,700 --> 00:05:08,683
Now it's your turn.

72
00:05:09,156 --> 00:05:13,671
Pick a sentence, follow its citation
to the row, and the row to the page.

73
00:05:15,234 --> 00:05:19,120
Here's the water figure: row E0159.

74
00:05:19,170 --> 00:05:20,883
Page 2, Table 1:

75
00:05:21,254 --> 00:05:25,179
−25.7%, in the Water use column.

76
00:05:26,619 --> 00:05:29,636
Do that for every figure
your conclusions rest on.

77
00:05:31,700 --> 00:05:33,945
The agents also listed
what they couldn't do:

78
00:05:34,457 --> 00:05:38,526
numbers shown only in charts, and
key papers missing from the folder.

79
00:05:39,606 --> 00:05:43,957
This run took 17 and a half minutes.
That's one example,

80
00:05:44,007 --> 00:05:47,415
on one folder.

81
00:05:50,900 --> 00:05:54,382
Pause now, and try the practice
folder from the companion page.

82
00:05:54,873 --> 00:05:58,545
Start your agent in the mode that
asks first, and paste the request.

83
00:05:58,956 --> 00:06:02,138
You should get a table
where every row names its page.

84
00:06:05,200 --> 00:06:08,925
So: write the rules,
give each agent one job,

85
00:06:09,276 --> 00:06:12,439
check against the source,
and follow the trail yourself.

86
00:06:14,900 --> 00:06:15,923
The rules file,

87
00:06:16,014 --> 00:06:20,250
the job descriptions and the practice
folder are all on the companion page.