Latest / Silicon & Shale Podcast / Episode 4: Eight Years in Two Weeks — AI R&D, the Hormuz Squeeze, and What Landmen Are Actually Afraid Of

Transcript

Episode 4: Eight Years in Two Weeks — AI R&D, the Hormuz Squeeze, and What Landmen Are Actually Afraid Of · Silicon & Shale Podcast

0:00
SPEAKER_00: Eight years.
0:01
SPEAKER_01: Eight years.
0:02
SPEAKER_00: An AI agent ran eight years of battery materials, R&D, in two weeks.
0:07
SPEAKER_00: Two weeks, Derek.
0:08
SPEAKER_01: And I keep trying to find the catch.
0:10
SPEAKER_01: I keep reading it looking for the asterisk.
0:13
SPEAKER_01: Eight years of R&D in two weeks.
0:16
And the asterisk says under ideal lab conditions or for a very
0:19
narrow problem or
0:21
SPEAKER_01: something that lets me put it in a box.
0:23
SPEAKER_00: There's no asterisk.
0:25
SPEAKER_01: There's no asterisk.
0:27
And that happened the same week the Strait of Hormuz went from
0:30
theoretical to actual.
0:32
SPEAKER_01: Tanker traffic dried up,
0:33
and LNG prices started doing things that should not be possible
0:37
this fast.
0:38
SPEAKER_00: Two completely different stories.
0:40
SPEAKER_00: One is about how fast the future is arriving.
0:42
SPEAKER_00: The other is about how the present just got a lot more complicated.
0:46
SPEAKER_01: And somehow this industry is living in both simultaneously.
0:49
Which is either the most interesting time to be in oil and gas,
0:52
or the most exhausting.
0:54
SPEAKER_01: Both.
0:54
SPEAKER_01: Definitely both.
0:56
SPEAKER_01: Welcome to Silicon and Shale,
0:58
SPEAKER_01: where the oil patch meets the algorithm,
1:01
SPEAKER_01: week of March 16th,
1:01
SPEAKER_01: 2026.
1:01
SPEAKER_01: Episode 4, People.
1:06
SPEAKER_00: We've made it to episode four.
1:08
SPEAKER_00: This week,
1:08
we've got the Strait of Hormuz actually doing the thing
1:11
everyone said it would do,
1:12
SPEAKER_00: what that means for markets.
1:14
And then we're going deep on an AI story that genuinely knocked
1:17
me sideways.
1:18
SPEAKER_00: Eight years of battery R&D done in two weeks by an AI agent.
1:23
SPEAKER_00: And then the collision zone,
1:24
which I'm very excited about because I think we landed on
1:27
something real this week.
1:28
SPEAKER_01: She's been sending me voice memos about it since 6 this morning.
1:32
SPEAKER_00: I had thoughts.
1:33
SPEAKER_01: She had thoughts.
1:34
SPEAKER_01: Let's get into it.
1:35
SPEAKER_01: All right, so last week, and look, we talked about this.
1:39
SPEAKER_01: We covered the U.S.
1:40
SPEAKER_01: strikes on Iran.
1:41
SPEAKER_01: Markets barely moved.
1:42
SPEAKER_01: We both kind of shrugged.
1:44
SPEAKER_01: The Strait of Hormuz thing felt theoretical.
1:46
SPEAKER_00: It is no longer theoretical.
1:48
SPEAKER_01: It is no longer theoretical.
1:50
SPEAKER_01: Tanker traffic has dried up.
1:52
SPEAKER_01: That's not hyperbole.
1:53
SPEAKER_01: That's the actual situation.
1:56
SPEAKER_01: And the prices that are moving aren't just crude.
1:58
SPEAKER_01: It's LNG.
1:59
SPEAKER_01: It's fertilizer.
2:01
Which is a whole second-order effect that I don't think enough
2:04
people were modeling.
2:05
SPEAKER_00: The fertilizer angle is wild.
2:07
Because you're talking about ammonia production,
2:09
you're talking about food prices downstream.
2:11
SPEAKER_00: This isn't just a gasoline story.
2:14
SPEAKER_01: And the market reaction has been more muted than you'd expect.
2:37
SPEAKER_01: More muted than it should be, in my opinion.
2:40
SPEAKER_01: And I think part of that is the SPR story, which we touched on last week.
2:44
The IEA starting to look more like a price management tool than
2:48
an emergency reserve.
2:50
SPEAKER_01: When the market knows there's a release valve, it prices in less risk.
2:54
SPEAKER_00: Which is actually kind of terrifying if you think about it.
2:56
SPEAKER_00: Because you're depleting the emergency reserve for non-emergencies.
3:00
And then a real emergency hits and...
3:02
You've got a sack of cement and a prayer.
3:05
SPEAKER_00: You've got a sack of cement and a prayer, exactly.
3:07
SPEAKER_01: Which,
3:08
SPEAKER_01: for anyone who missed episode one,
3:10
SPEAKER_01: that's my description of standard Oklahoma well-abandonment protocol,
3:14
SPEAKER_01: but it applies broadly.
3:16
SPEAKER_00: It applies so broadly.
3:18
SPEAKER_00: But okay,
3:18
SPEAKER_00: the Jones Act waiver,
3:20
SPEAKER_00: because that's been floating around this week,
3:22
and I want to make sure we actually explain what that is for
3:24
people who don't live
3:25
SPEAKER_00: and breathe this stuff.
3:26
SPEAKER_01: The Jones Act says that shipping between U.S.
3:29
SPEAKER_01: ports has to happen on U.S.-flagged,
3:31
SPEAKER_01: U.S.-built,
3:32
SPEAKER_01: U.S.-crewed vessels,
3:34
SPEAKER_01: which sounds fine in theory.
3:36
SPEAKER_01: In practice,
3:37
SPEAKER_01: it means that if there's a supply disruption on the Gulf Coast,
3:40
SPEAKER_01: you can't just reroute a foreign tanker to fix it without a waiver.
3:44
SPEAKER_00: And waivers have been used more and more as a policy tool.
3:47
Which again starts to feel less like emergency management and
3:50
more like market management.
3:52
SPEAKER_01: And those are different things with different risk profiles.
3:56
SPEAKER_00: The reservoir damage fears.
3:58
SPEAKER_00: There was some chatter about that this week.
4:00
SPEAKER_00: Crude Realities covered it and basically said it's mostly noise.
4:04
SPEAKER_01: Mostly noise.
4:06
The idea that you do meaningful reservoir damage in a conflict
4:09
scenario is,
4:10
SPEAKER_01: look,
4:11
SPEAKER_01: it's not zero,
4:12
but it's a lot harder to permanently damage a producing
4:15
reservoir than people
4:16
SPEAKER_01: think.
4:17
SPEAKER_01: The Kuwaiti fields in 91 are the reference point everyone uses,
4:22
SPEAKER_01: and even those came back faster than the apocalypse crowd predicted.
4:26
SPEAKER_00: Derek just casually dropping the Kuwait war as his reference point.
4:30
SPEAKER_00: This is why I need him.
4:32
SPEAKER_01: 25 years in this business, you accumulate reference points.
4:35
SPEAKER_00: Okay, but here's my question.
4:37
SPEAKER_00: And this is where the data management angle comes in,
4:40
SPEAKER_00: because there was a really interesting episode this week on upstream data
4:44
SPEAKER_00: challenges.
4:45
SPEAKER_00: When we're talking about disruption of this scale,
4:48
how much of the response capacity depends on operators actually
4:51
knowing what they
4:52
SPEAKER_00: have?
4:53
SPEAKER_01: Oh, that's a good question.
4:54
SPEAKER_01: Say more.
4:55
SPEAKER_00: There was a conversation with a data management consultant,
4:58
SPEAKER_00: or TECHA,
4:59
SPEAKER_00: they do industrial data work.
5:01
SPEAKER_00: And the core point was that data quality in oil and gas is still,
5:05
SPEAKER_00: even in 2026,
5:07
SPEAKER_00: a massive unsolved problem.
5:09
SPEAKER_00: Not a technology problem, a culture and process problem.
5:14
And if you're an operator trying to respond quickly to a supply
5:17
shock and your
5:18
SPEAKER_00: asset data is garbage,
5:19
SPEAKER_01: then you're back to the binder.
5:21
SPEAKER_00: Then you're back to the binder.
5:23
SPEAKER_00: And I think that's underappreciated in the geopolitical conversation.
5:28
SPEAKER_00: Everyone talks about spare capacity.
5:30
Nobody talks about whether you can actually activate spare
5:33
capacity quickly when
5:34
your data infrastructure is held together with spreadsheets and
5:37
institutional
5:37
SPEAKER_00: memory.
5:38
SPEAKER_01: That is, okay, that's genuinely a good point.
5:42
SPEAKER_01: Because I've seen this,
5:43
I've been on projects where an operator is trying to ramp up
5:46
production and they
5:47
spend the first six weeks just figuring out what they actually
5:50
own and what the
5:51
SPEAKER_01: status of it is.
5:52
SPEAKER_00: And that's the metadata problem.
5:55
SPEAKER_00: It's not even the data itself.
5:56
It's knowing what the data means,
5:57
where it came from,
5:59
whether you can trust it.
6:00
SPEAKER_01: The data quality thing in this industry is real.
6:03
SPEAKER_01: And I'll tell you, on the land side, it's arguably worse.
6:07
SPEAKER_01: You've got lease records,
6:08
SPEAKER_01: division order titles,
6:09
SPEAKER_01: production allocations,
6:11
SPEAKER_01: and a lot of that still lives in systems that don't talk to each other.
6:14
SPEAKER_00: Or in binders.
6:16
SPEAKER_01: Or in binders in Tillman County.
6:18
SPEAKER_00: Okay,
6:19
SPEAKER_00: speaking of things that are genuinely impressive right now,
6:23
SPEAKER_00: the AI story that's been bouncing around my brain all week.
6:26
SPEAKER_01: The battery one.
6:27
SPEAKER_00: The battery one.
6:29
SPEAKER_00: So SES AI, they make lithium metal batteries.
6:33
These are next generation batteries,
6:35
way more energy dense than what's in your phone right now.
6:39
Their CEO talked about how they used AI agents to compress what
6:42
would have been an
6:43
SPEAKER_00: eight-year R&D cycle into two weeks.
6:45
SPEAKER_01: Two weeks.
6:46
SPEAKER_00: Two weeks.
6:47
And I need to explain what that actually means because it's
6:50
easy to just say,
6:51
SPEAKER_00: AI did science faster and not understand the mechanism.
6:55
SPEAKER_00: These agents were running what they call wet lab experiments,
7:00
SPEAKER_00: actual physical experiments,
7:02
guided by AI models that were iterating on what to try next
7:05
based on what they'd
7:06
SPEAKER_00: already learned.
7:07
SPEAKER_01: So it's not just AI analyzing existing data.
7:10
SPEAKER_00: It's AI designing the experiments,
7:12
SPEAKER_00: running them,
7:13
SPEAKER_00: reading the results,
7:14
SPEAKER_00: and deciding what to try next.
7:16
SPEAKER_00: autonomously, in a loop.
7:19
SPEAKER_00: And the domain is molecular chemistry,
7:21
SPEAKER_00: which is,
7:22
SPEAKER_00: it's a combinatorial search space that's essentially infinite.
7:26
There are more possible molecular configurations than there are
7:29
atoms in the
7:30
SPEAKER_00: observable universe.
7:32
Okay,
7:32
I'm going to ask the dumb question because I think it's
7:34
actually important.
7:36
SPEAKER_01: When they say eight years into two weeks,
7:39
SPEAKER_01: is that wall clock time or is that human hours compressed?
7:42
SPEAKER_00: Great question.
7:43
SPEAKER_00: It's both, actually.
7:45
SPEAKER_00: The AI runs the experimental loop continuously.
7:48
No sleep,
7:49
no conferences,
7:50
no waiting for the grad student to get back from lunch.
7:54
SPEAKER_00: But also the search is smarter.
7:56
SPEAKER_00: It's not random.
7:57
SPEAKER_00: It's guided by the model's understanding of what's likely to work.
8:01
So it's faster in two ways,
8:03
the pace of work and the quality of the search.
8:05
SPEAKER_00: Exactly.
8:06
And the implications for upstream,
8:09
because we always have to bring it back.
8:11
SPEAKER_01: Always.
8:12
SPEAKER_00: The upstream parallel is fluid characterization.
8:16
SPEAKER_00: Enhanced recovery chemistry.
8:17
You've got operators running EOR programs where they're trying
8:21
to figure out what
8:22
SPEAKER_00: surfactant blend works best in a specific reservoir.
8:25
SPEAKER_00: And that's fundamentally the same kind of problem.
8:28
SPEAKER_00: Combinatorial chemistry search with a physical experiment at the end.
8:32
SPEAKER_01: That's not nothing.
8:33
SPEAKER_01: Because EOR optimization right now is largely empirical.
8:37
SPEAKER_01: You run a pilot, you wait, you see what happens, you adjust.
8:41
What if the AI is designing the next pilot based on the last
8:45
one's results before
8:46
SPEAKER_00: your engineering team has even had their debrief meeting?
8:49
SPEAKER_01: That actually gives me a slightly uncomfortable feeling,
8:51
SPEAKER_01: and I'm trying to figure out if it's excitement or vertigo.
8:54
SPEAKER_00: Probably both.
8:56
SPEAKER_00: Same energy as the binder, really.
8:58
SPEAKER_01: Don't compare the binder to AI discovery.
9:00
SPEAKER_01: The binder is a trauma response.
9:02
SPEAKER_00: Okay.
9:03
SPEAKER_00: The other AI story I want to hit,
9:06
SPEAKER_00: and this one is more in the,
9:07
SPEAKER_00: oh,
9:07
SPEAKER_00: this is real and it's happening now category.
9:11
SPEAKER_00: is this whole anthropic story.
9:13
Because last week we covered the Pentagon deal and the
9:15
blacklisting question.
9:16
SPEAKER_00: And it's escalating.
9:18
SPEAKER_01: The Can AI Handle 94% of KnowledgeWorks Study?
9:22
SPEAKER_00: That one.
9:24
SPEAKER_00: Anthropic put out what they're calling an observed exposure study.
9:28
The claim is that AI can theoretically handle 94% of knowledge
9:32
work tasks.
9:33
SPEAKER_00: And the word theoretically is doing a lot of work in that sentence.
9:37
SPEAKER_00: But the gap between theoretical and actual is narrowing.
9:40
SPEAKER_01: How fast is it narrowing?
9:42
SPEAKER_00: That's the question nobody has a clean answer to.
9:45
SPEAKER_00: But here's the thing that stuck with me.
9:48
A Sequoia partner was quoted saying this week that AI isn't
9:51
just replacing
9:52
SPEAKER_00: software,
9:53
SPEAKER_00: it's replacing entire service industries,
9:55
SPEAKER_00: not automating tasks within a service,
9:58
SPEAKER_00: replacing the service.
10:00
SPEAKER_01: What does that mean for land services?
10:02
SPEAKER_01: Because that's my world.
10:04
SPEAKER_00: It means, okay, think about the stuff that landmen currently outsource.
10:09
Title abstracting,
10:10
lease analysis,
10:12
curative work,
10:13
due diligence on acquisitions,
10:15
SPEAKER_01: All of that is high-skill, judgment-heavy work.
10:18
SPEAKER_00: All of it is also, at some level, pattern recognition on documents.
10:23
And pattern recognition on documents is exactly what these
10:26
models are getting very good at.
10:28
SPEAKER_01: I'm going to push back slightly,
10:29
SPEAKER_01: though,
10:29
SPEAKER_01: and I want to be precise about where,
10:32
because there's a version of this conversation that undersells
10:35
the judgment piece.
10:36
When I'm running a title chain and I find a gap,
10:39
let me give you the specific case.
10:41
SPEAKER_01: There's a 1962 mineral deed out of Tillman County,
10:45
SPEAKER_01: same county,
10:45
SPEAKER_01: actually,
10:46
SPEAKER_01: where the grantor description is ambiguous.
10:49
SPEAKER_01: It could be the same party as the grantee two links up,
10:53
SPEAKER_01: or it could be a different person with a similar name.
10:55
SPEAKER_00: And a human reads the surrounding context and makes a call.
10:58
SPEAKER_01: A human reads the surrounding context,
11:01
SPEAKER_01: calls the county assessor,
11:03
SPEAKER_01: checks the obituaries,
11:04
SPEAKER_01: sometimes literally drives to the courthouse,
11:07
SPEAKER_01: and looks at the original instrument because the scan is bad.
11:10
SPEAKER_01: That's not pattern matching.
11:12
SPEAKER_01: That's investigation.
11:14
SPEAKER_00: Okay, but... And I mean this genuinely, not dismissively.
11:18
How much of what you just described is the AI actually can't do
11:22
versus the AI
11:23
SPEAKER_00: hasn't been trained to do in this specific domain?
11:26
SPEAKER_01: That's a fair distinction.
11:27
SPEAKER_00: Because the multi-step investigation loop you just described,
11:30
where you read something ambiguous and then go find more
11:33
information and then
11:34
SPEAKER_00: synthesize,
11:35
SPEAKER_00: that is exactly what agents do now.
11:38
Fax machine to fully autonomous on AI handling that tidal
11:41
ambiguity problem.
11:42
SPEAKER_01: Where do we land?
11:43
SPEAKER_00: I'm going to say five.
11:45
SPEAKER_00: Not because the capability isn't there in theory,
11:48
SPEAKER_00: but because the specific integration,
11:50
SPEAKER_00: courthouse databases,
11:52
SPEAKER_00: county-level record systems,
11:53
SPEAKER_00: half of which are not digitized.
11:56
SPEAKER_00: The data layer isn't ready for the model to work on.
11:58
SPEAKER_01: I'll give it a four.
12:00
SPEAKER_01: And that's actually higher than I would have said six months ago.
12:03
SPEAKER_00: Progress.
12:04
SPEAKER_00: Genuine progress.
12:05
SPEAKER_01: Don't make it a thing.
12:06
SPEAKER_00: Too late.
12:07
SPEAKER_00: I'm noting it.
12:08
SPEAKER_01: Okay, collision zone.
12:09
SPEAKER_00: Collision zone.
12:11
SPEAKER_01: So here's the problem I want to start with,
12:13
SPEAKER_01: and this is something I've been annoyed about for my entire career.
12:17
SPEAKER_01: When there's a supply shock,
12:19
SPEAKER_01: a real one,
12:20
SPEAKER_01: like what's happening right now in the straight,
12:22
SPEAKER_01: the response time for domestic production is measured in months,
12:26
SPEAKER_01: not days,
12:27
SPEAKER_01: months.
12:28
SPEAKER_01: And a big piece of why is what we were talking about earlier.
12:31
Operators don't have a clean,
12:33
real-time picture of their own inventory of ready-but-idle
12:36
wells.
12:37
SPEAKER_00: The drilled but uncompleted wells, DUCs, and shut-in production.
12:42
SPEAKER_01: DUCs,
12:43
SPEAKER_01: shut-ins,
12:44
wells on temporary abandonment that could be returned to
12:48
production with the right
12:49
SPEAKER_01: economics.
12:50
There's a whole shadow inventory of potential production that
12:53
nobody has a great
12:54
SPEAKER_01: handle on industry-wide.
12:56
SPEAKER_00: And even at the operator level, the picture is fragmented.
12:59
SPEAKER_01: At the operator level,
13:00
it's fragmented because the data lives in three different
13:03
systems that don't talk
13:04
SPEAKER_01: to each other,
13:05
SPEAKER_01: plus a spreadsheet that one guy maintains,
13:08
plus institutional knowledge held by a reservoir engineer who's
13:11
been there since
13:12
SPEAKER_01: 2008 and hasn't documented any of it.
13:15
SPEAKER_00: Okay.
13:16
So this is the data management problem with a geopolitical
13:19
urgency attached to it.
13:20
SPEAKER_01: Exactly.
13:22
And it's been tolerable in a normal market because you have
13:24
time to figure it out.
13:26
SPEAKER_01: In a shock scenario, you don't.
13:28
So here's where I think this connects to the AI stuff we were
13:31
talking about.
13:32
SPEAKER_00: The research compression story, the eight years in two weeks thing.
13:37
SPEAKER_00: The core mechanism is an agent that runs a loop.
13:40
Gather information,
13:41
decide what to try,
13:42
run the experiment,
13:45
read the results.
13:47
SPEAKER_00: Repeat, that loop is not unique to battery chemistry.
13:50
SPEAKER_01: You're saying you could run that loop on well inventory data.
13:53
SPEAKER_00: I'm saying imagine a system that continuously ingests production data,
13:57
SPEAKER_00: completion records,
13:59
SPEAKER_00: regulatory filings,
14:00
SPEAKER_00: land records,
14:02
SPEAKER_00: which I know is the hard part,
14:03
SPEAKER_00: Derek.
14:04
SPEAKER_00: Don't make a face.
14:05
SPEAKER_01: I'm not making a face.
14:07
SPEAKER_00: You're making the face.
14:09
SPEAKER_00: But imagine it knows,
14:10
SPEAKER_00: at any given moment,
14:12
here are the hundred wells in this basin that could be brought
14:14
back to production
14:15
SPEAKER_00: fastest.
14:17
SPEAKER_00: Here's what it would cost.
14:19
SPEAKER_00: Here's what the expected rate would be.
14:21
SPEAKER_00: Here's the lease status on each one.
14:23
SPEAKER_01: And that last one is where it falls down.
14:26
Because lease status isn't a clean field in a database,
14:29
it's a chain of title question.
14:32
It's whether the primary term has expired and whether there's a
14:35
savings clause and
14:36
whether production in paying quantities was established before
14:39
the expiration date.
14:40
SPEAKER_00: Which is why the output of this system isn't push-button, start wells.
14:45
SPEAKER_00: It's, here are the 10 wells most worth sending Derek to look at first.
14:49
SPEAKER_01: Okay.
14:51
SPEAKER_01: Okay, that framing I can work with.
14:53
Because the problem right now is that nobody knows where to
14:55
point the human attention.
14:57
SPEAKER_01: There's too much to look at.
14:59
SPEAKER_00: The AI is the triage layer, not the decision maker.
15:03
SPEAKER_01: The AI is the triage layer.
15:05
SPEAKER_01: It surfaces what's worth the investigation.
15:08
And then a landman or a production engineer or whoever applies
15:11
the judgment.
15:12
SPEAKER_00: And in a shock scenario,
15:14
a real one where you've got 90 days to show Congress that
15:17
domestic production is
15:18
SPEAKER_00: responding,
15:19
that triage layer could mean the difference between actually
15:22
responding and just
15:23
SPEAKER_00: putting out a press release.
15:24
SPEAKER_01: You know what?
15:25
That's one of the more concrete cases I've heard for why this
15:28
stuff matters in the field.
15:30
Not AI will do your job,
15:32
but AI will tell you which jobs actually need to be done right
15:35
now.
15:36
SPEAKER_00: Prioritization at scale.
15:37
SPEAKER_00: That's the whole thing.
15:39
SPEAKER_01: And the data management piece we were talking about earlier,
15:41
SPEAKER_01: the Orteca conversation,
15:43
SPEAKER_01: the metadata problem,
15:45
SPEAKER_01: that's the prerequisite.
15:46
SPEAKER_01: You can't run the system on garbage data.
15:49
SPEAKER_01: You need the data foundation first.
15:51
SPEAKER_00: Which is why,
15:51
SPEAKER_00: honestly,
15:52
the most valuable thing a lot of operators could do right now
15:55
is not buy the AI
15:56
SPEAKER_00: tool.
15:57
SPEAKER_00: It's clean up your data so the AI tool has something to work with.
16:01
That's the unsexy version of the pitch that never makes it onto
16:04
the conference slide.
16:05
SPEAKER_00: Because clean your data doesn't have a keynote demo.
16:08
SPEAKER_01: But it's the actual answer.
16:10
SPEAKER_00: It's the actual answer.
16:11
SPEAKER_00: Every time.
16:12
SPEAKER_01: All right, build list.
16:14
SPEAKER_01: What are we building?
16:14
SPEAKER_00: Okay.
16:16
SPEAKER_00: So number one this week is a readiness intelligence layer for operators.
16:20
SPEAKER_00: I'm not going to call it a DUC tracker because that's too narrow.
16:24
SPEAKER_00: Think of it as a can we respond dashboard.
16:28
At any moment,
16:29
what's the realistic upside production capacity in your
16:32
operated portfolio?
16:33
SPEAKER_00: What are the constraints on each well or pad?
16:35
SPEAKER_00: And what's the ranking by speed to production?
16:38
SPEAKER_01: The land piece of that,
16:39
SPEAKER_01: the lease status,
16:40
SPEAKER_01: the surface access,
16:41
SPEAKER_01: the regulatory standing,
16:43
has never been integrated with the production and engineering
16:45
data in a way that's
16:46
SPEAKER_01: actually useful to someone making a fast decision.
16:50
SPEAKER_00: And that integration is the thing worth building.
16:53
SPEAKER_00: Not any one piece of it.
16:54
SPEAKER_00: The production data tools exist.
16:57
SPEAKER_00: The land management tools exist.
17:00
It's the connection layer that synthesizes them into a single
17:03
readiness picture.
17:04
The second thing I'd put on this list is the geopolitical early
17:07
warning layer for
17:08
SPEAKER_01: procurement,
17:09
SPEAKER_01: because the fertilizer price story hit me this week.
17:13
SPEAKER_01: If you're an operator with significant chemical exposure,
17:17
SPEAKER_01: work over chemicals,
17:18
SPEAKER_01: completion fluids,
17:19
SPEAKER_01: anything that's got ammonia-derived components,
17:22
the Hormuz story has a supply chain tale that almost nobody is
17:25
tracking in real
17:26
SPEAKER_01: time.
17:27
A system that connects geopolitical signals to your specific
17:30
supply chain exposure.
17:32
SPEAKER_00: Like,
17:33
SPEAKER_00: this tanker route disruption affects this category of chemicals,
17:36
SPEAKER_00: which affects these upcoming completion jobs by this estimated amount.
17:40
SPEAKER_01: Right now,
17:40
SPEAKER_01: that analysis happens in someone's head,
17:43
SPEAKER_01: maybe,
17:43
SPEAKER_01: after they've already gotten the price shock call from their supplier.
17:47
SPEAKER_00: It should happen before the call.
17:48
SPEAKER_00: And the tools to build that exist now.
17:51
SPEAKER_01: They do.
17:52
SPEAKER_00: OK, that's midnight build list items.
17:55
SPEAKER_00: I've genuinely lost count.
17:57
SPEAKER_00: I think we're at 16 at this point.
17:59
Someone in our audience is definitely going to build one of
18:02
these before I do.
18:04
SPEAKER_01: That's fine.
18:05
SPEAKER_01: Build it.
18:05
SPEAKER_00: Build it and tell us.
18:07
SPEAKER_00: And if you're listening to this thinking,
18:09
SPEAKER_00: I actually need that built for my operation,
18:11
SPEAKER_00: reach out.
18:12
SPEAKER_00: Siliconandshaleagmail.com.
18:15
SPEAKER_00: We build this stuff.
18:16
SPEAKER_01: We do.
18:16
SPEAKER_01: That's not hypothetical.
18:18
SPEAKER_00: The readiness intelligence thing especially.
18:21
SPEAKER_00: That one is overdue.
18:22
SPEAKER_01: Overdue by about 15 years.
18:24
SPEAKER_00: Okay, look.
18:26
SPEAKER_00: I want to come back to something before we wrap up.
18:29
SPEAKER_00: The AI can handle 94% of knowledge work stat.
18:33
Because I've been sitting with it all week and I can't decide
18:35
if it's a threat or a promise.
18:37
SPEAKER_01: What's the difference?
18:39
SPEAKER_00: Whether you're the one who gets to decide what happens with the other 6%.
18:42
SPEAKER_00: That's actually... Because the 6% is the stuff that matters.
18:47
SPEAKER_00: The ambiguous deed.
18:48
SPEAKER_00: The judgment call under pressure.
18:51
The relationship with the county clerk that gets you access to
18:53
the index binder
18:54
SPEAKER_00: when the computer's been down since Tuesday.
18:57
SPEAKER_01: I want that on a bumper sticker.
18:58
SPEAKER_01: The 6% is the stuff that matters.
19:01
SPEAKER_00: The 6% is where all of the value lives.
19:04
SPEAKER_00: And if you've spent your career building that 6% expertise,
19:08
SPEAKER_00: actually building it,
19:09
SPEAKER_00: not just showing up to the job,
19:11
SPEAKER_00: then this moment is not as scary as the headline makes it sound.
19:15
If you're a generalist who can do everything at a mediocre
19:17
level,
19:18
that's the hard conversation.
19:20
SPEAKER_00: That's the hard conversation.
19:22
But if you're Derek and you can read a 1947 handwritten deed
19:25
and figure out what it
19:26
SPEAKER_00: actually means in 2026,
19:28
SPEAKER_00: no model is replacing that tomorrow.
19:30
SPEAKER_01: I appreciate you saying that.
19:33
I'm going to go stand in more county basements and feel
19:35
slightly better about my career.
19:38
And if any of you out there sell software to operators or
19:40
landmen and you want to
19:41
SPEAKER_00: get in front of the people who actually buy it,
19:44
SPEAKER_00: we're taking sponsors,
19:46
SPEAKER_00: siliconandshaleegmail.com.
19:48
SPEAKER_00: We'll actually use your product before we talk about it,
19:50
SPEAKER_00: which is more than most podcasts can say.
19:53
SPEAKER_01: That is true.
19:53
SPEAKER_01: We will.
19:54
SPEAKER_00: Okay, I actually have to go.
19:56
I have a data pipeline that has been broken since yesterday,
19:59
and I've been pretending it's fine.
20:01
SPEAKER_01: Is it fine?
20:02
SPEAKER_00: It is not fine.
20:04
It's very much the pipeline equivalent of the computer being
20:06
down since Tuesday.
20:08
SPEAKER_01: The fax machine is still warm, Pac.
20:10
SPEAKER_00: Go away.