1
00:00:00,000 --> 00:00:03,764
Sam: Do I need to buy runners or move my code into a hosted service?

2
00:00:03,917 --> 00:00:05,711
LitAI: No new fleet is required.

3
00:00:05,875 --> 00:00:09,879
LitAI: This course connects machines you already own to Literate AI as
workers.

4
00:00:10,042 --> 00:00:17,403
LitAI: The key boundary is simple: the project declares what it needs,
while your private user configuration chooses where work runs.

5
00:00:17,553 --> 00:00:22,300
LitAI: These are edited transcript replays of real sessions, not raw
screen recordings.

6
00:00:22,470 --> 00:00:24,601
LitAI: Remote-worker steps are a walkthrough.

7
00:00:24,761 --> 00:00:28,571
Sam: Why not just put the SSH host in my project file?

8
00:00:28,721 --> 00:00:34,357
LitAI: Because endpoints and credentials are private execution
choices, not portable product authority.

9
00:00:34,513 --> 00:00:37,006
LitAI: A target and a worker are different decisions.

10
00:00:37,156 --> 00:00:44,935
LitAI: The target determines the recipe and intended platform; the
worker is the local, command, or SSH endpoint that executes it.

11
00:00:45,085 --> 00:00:48,819
LitAI: Never encode private hostnames in project specifications.

12
00:00:48,969 --> 00:00:53,331
LitAI: Ask litai config paths where this platform stores user
configuration.

13
00:00:53,481 --> 00:00:59,796
LitAI: Copy the checked-in synthetic examples to those private
locations, or pass explicit absolute paths.

14
00:00:59,946 --> 00:01:03,061
LitAI: Do not guess Windows or Unix conventions from memory.

15
00:01:03,211 --> 00:01:08,851
LitAI: In your private workers file, replace each example endpoint and
workspace with a real assignment.

16
00:01:09,003 --> 00:01:13,233
LitAI: Give each machine a stable worker ID and exact platform
requirements.

17
00:01:13,383 --> 00:01:16,063
LitAI: Keep credentials out of this JSON file.

18
00:01:16,216 --> 00:01:20,280
LitAI: The test matrix selects workers by ID and matching platform
Flavor.

19
00:01:20,430 --> 00:01:23,852
LitAI: It also selects a coding CLI and model for live generation.

20
00:01:24,013 --> 00:01:28,286
LitAI: The matrix does not repeat SSH credentials or provisioning
commands.

21
00:01:28,436 --> 00:01:30,378
Sam: My model name works in chat.

22
00:01:30,528 --> 00:01:32,120
Sam: Why verify it again?

23
00:01:32,278 --> 00:01:36,635
LitAI: Verify the model before a long run, then probe every configured
worker.

24
00:01:36,786 --> 00:01:46,070
LitAI: A missing NVIDIA utility means no GPU was discovered; a utility
that cannot query its driver is a degraded observation, not proof of
zero devices.

25
00:01:46,220 --> 00:01:49,817
LitAI: Prepare each assigned machine with detect-before-install
checks.

26
00:01:49,970 --> 00:01:55,261
LitAI: Hydrate the exact Git revision, including required submodules
and LFS content.

27
00:01:55,429 --> 00:02:00,686
LitAI: Coding-agent login is a separate readiness gate, and secrets
stay in the worker environment.

28
00:02:00,845 --> 00:02:03,413
LitAI: Start with one sample and one worker cell.

29
00:02:03,563 --> 00:02:09,454
LitAI: Only expand the matrix after platform, source authority, and
toolchain preflight pass.

30
00:02:09,605 --> 00:02:15,028
LitAI: For this repository, samples-platform-regression is the
supported fan-out demonstration.

31
00:02:15,188 --> 00:02:18,825
Sam: If the remote machine says it passed, why isn't that enough?

32
00:02:18,980 --> 00:02:24,531
LitAI: For a single Component, build, test, and run can choose a
configured worker by ID.

33
00:02:24,688 --> 00:02:31,836
LitAI: Returned artifacts and receipts must be imported and verified
by the coordinator; a path on another machine alone is not evidence.

34
00:02:31,986 --> 00:02:36,301
LitAI: Our real package build used the implicit local worker and
produced a pip wheel.

35
00:02:36,451 --> 00:02:40,873
LitAI: That verifies the local execution path without exposing a host
address.

36
00:02:41,034 --> 00:02:48,093
LitAI: To demonstrate an SSH worker, configure one of your own
machines privately and rerun the same target with its worker ID.

37
00:02:48,243 --> 00:02:52,699
LitAI: This can make cross-platform tests more practical without
buying a new service.

38
00:02:52,868 --> 00:02:57,883
LitAI: Capacity probes inform selection, but never rewrite authored
minimum requirements.

39
00:02:58,035 --> 00:03:05,217
LitAI: Derived projects can wire the private matrix into their own CI;
the packaged CLI does not promise a generic fan-out verb.