Latest / Silicon & Shale Podcast / Episode 11: The week’s E&P developments & applied AI

Transcript

Episode 11: The week’s E&P developments & applied AI · Silicon & Shale Podcast

0:00
SPEAKER_00: So Amazon's apparently building a 7.6 gigawatt private gas plant in West Texas.
0:04
SPEAKER_00: I don't even know where to start with that number.
0:05
SPEAKER_01: Wait, 7.6?
0:07
SPEAKER_01: As in singular facility?
0:08
SPEAKER_00: What's being reported is a private gas grid, not just a plant.
0:12
SPEAKER_00: They're basically standing up their own utility-scale power infrastructure to feed data centers.
0:17
SPEAKER_00: I've seen operators build their own gas processing,
0:19
SPEAKER_00: their own water infrastructure, but this is a different animal entirely.
0:22
SPEAKER_01: That's roughly the output of seven or eight nuclear reactors.
0:25
SPEAKER_01: For one company's compute demand.
0:27
SPEAKER_01: And they're doing it in the Permian,
0:28
SPEAKER_01: which means they're competing for pipeline capacity, water, labor, pad sites.
0:33
SPEAKER_01: Everything operators are already fighting over.
0:35
SPEAKER_00: The part nobody's talking about is what it does to the land market.
0:38
SPEAKER_00: You drop something that size into the basin, you're not just consuming gas.
0:41
SPEAKER_00: You're reshaping surface economics.
0:43
SPEAKER_00: You're pulling crews.
0:44
SPEAKER_00: You're tying up easements.
0:45
SPEAKER_00: The surface use agreements alone for something at that scale would be a nightmare.
0:49
SPEAKER_00: I've seen a single 200-megawatt facility gum up a county's permitting office for months.
0:53
SPEAKER_01: So what happens when the landman shows up and the surface owner is already leased to
0:56
SPEAKER_01: a tech company for a substation?
0:58
SPEAKER_01: That's a collision nobody's modeling yet.
1:00
SPEAKER_00: Some people are.
1:01
SPEAKER_00: Quietly.
1:02
SPEAKER_00: I talked to a guy running title in Reeves County last month
1:04
SPEAKER_00: who said he'd already seen two tracks where
1:06
SPEAKER_00: the surface rights had fresh encumbrances from an entity he'd never heard of,
1:09
SPEAKER_00: Shell Company.
1:10
SPEAKER_00: Couldn't trace it back to anything obvious,
1:11
SPEAKER_00: but the acreage was right along a major transmission corridor.
1:14
SPEAKER_01: We might already be looking at a land grab that hasn't made the headlines.
1:17
SPEAKER_01: Shell entities tying up surface before operators even know what's coming.
1:21
SPEAKER_00: And the mineral owners underneath have no idea their surface access just got
1:24
SPEAKER_00: a lot more complicated.
1:25
SPEAKER_00: That's the thing that keeps me up, not the gigawatts.
1:28
SPEAKER_01: Welcome to Silicon and Shale, where the oil patch meets the algorithm.
1:32
SPEAKER_01: It's the week of August 18th, 2026, episode 12.
1:36
SPEAKER_01: And we're already in it,
1:37
SPEAKER_01: so if you just tuned in,
1:38
SPEAKER_01: we're talking about Amazon potentially building a private gas grid
1:41
SPEAKER_01: in West Texas that's bigger than some countries' entire power systems,
1:44
SPEAKER_01: and what that means for the people who actually work the land out there.
1:47
SPEAKER_00: The question isn't how many gigawatts or what it costs Amazon.
1:50
SPEAKER_00: The question is what happens to the operator who shows up Tuesday morning
1:53
SPEAKER_00: and finds out the Surface game changed while he was running Tidal.
1:56
SPEAKER_01: So today we're going deep on that collision.
1:58
SPEAKER_01: Tech infrastructure versus E&P access, who wins when they're competing for the same dirt.
2:03
SPEAKER_01: And then later,
2:03
SPEAKER_01: the Gulf lease sale that just pulled in over 80 million in high bids,
2:07
SPEAKER_01: what the NPC is saying about critical minerals,
2:09
SPEAKER_01: and whether any of this AI agent hype actually survives contact with
2:11
SPEAKER_01: a real asset team.
2:13
SPEAKER_01: Twelve episodes in, Derek.
2:15
SPEAKER_01: We're still here.
2:16
SPEAKER_00: Lower the bar pack, where still here is about right.
2:18
SPEAKER_00: But here's what actually matters this week on the E&P side, the Gulf lease sale.
2:22
SPEAKER_00: 82.7 million in high bids, 16 companies showing up.
2:26
SPEAKER_00: That's not a soft market.
2:27
SPEAKER_01: Deepwater, specifically.
2:29
SPEAKER_01: Not someone throwing a few hundred grand at a shallow prospect.
2:32
SPEAKER_01: Those are companies making billion-dollar development commitments based
2:34
SPEAKER_01: on what they think the next decade looks like.
2:36
SPEAKER_00: And you've got to remember the context.
2:38
SPEAKER_00: There's been all this noise about peak demand narratives,
2:40
SPEAKER_00: energy transition timelines,
2:42
SPEAKER_00: and yet 16 operators walked into that sale and put real money down on Gulf
2:46
SPEAKER_00: of Mexico acreage, which isn't exactly a low-cost basin.
2:49
SPEAKER_01: The composition matters more than the total number, though.
2:51
SPEAKER_01: Who showed up?
2:53
SPEAKER_00: If the usual suspects, your shells, your BPs, your chevrons, that tells one story.
2:57
SPEAKER_00: If you're seeing mid-caps or even private stepping into deep water,
3:00
SPEAKER_00: that tells you something completely different about where capital's flowing.
3:04
SPEAKER_01: The risk calculus shifted.
3:05
SPEAKER_01: The Gulf's been underappreciated for years.
3:07
SPEAKER_01: The infrastructure's there.
3:08
SPEAKER_01: The regulatory framework is, well, it's known.
3:11
SPEAKER_01: You can model it.
3:12
SPEAKER_01: Compared to trying to permit something onshore in half the states right now,
3:15
SPEAKER_01: the Gulf might actually be the path of least resistance for some of these operators.
3:18
SPEAKER_00: I wouldn't go that far.
3:20
SPEAKER_00: BSEE isn't handing out gold stars.
3:22
SPEAKER_00: But compared to fighting over surface access in the Permian while Amazon's building
3:25
SPEAKER_00: a power grid next door...
3:27
SPEAKER_01: You can see why someone might look at a deepwater block and think,
3:29
SPEAKER_01: at least I know what I'm dealing with.
3:31
SPEAKER_00: That's the calculation more people are running than they'll admit publicly.
3:34
SPEAKER_01: 82 million in high bids.
3:35
SPEAKER_01: How does that compare historically?
3:37
SPEAKER_01: Because people hear a number and don't have a baseline.
3:40
SPEAKER_00: Solid.
3:41
SPEAKER_00: Not 2019 numbers, but real.
3:43
SPEAKER_00: And the fact that it happened now,
3:44
SPEAKER_00: with crude sitting where it is, with all the Iran risk still baked in.
3:48
SPEAKER_00: Operators are telling you they think the commodity's got legs.
3:51
SPEAKER_00: You don't commit to deep water development if you think oil's going to 50.
3:54
SPEAKER_01: The cycle times alone make that a terrible bet at 50.
3:57
SPEAKER_01: Four, five years before first oil on some of these prospects.
4:00
SPEAKER_00: So they're looking at 2030-2031 economics when they write that check.
4:03
SPEAKER_01: Which brings up the NPC piece.
4:05
SPEAKER_01: John DeBar, he's the executive director, was on a podcast this week from OTC.
4:10
SPEAKER_01: And he was talking about something that doesn't get enough airtime.
4:12
SPEAKER_01: Critical minerals.
4:14
SPEAKER_00: Go on.
4:14
SPEAKER_01: The framing was interesting because he wasn't just talking about lithium and cobalt
4:17
SPEAKER_01: for batteries.
4:18
SPEAKER_01: He was talking about the minerals that oil and gas itself depends on.
4:21
SPEAKER_01: The stuff in your downhole tools, your drilling equipment, your subsea completions hardware.
4:25
SPEAKER_01: There's a supply chain vulnerability there that nobody in ENP wants to confront.
4:29
SPEAKER_00: Because nobody's writing a LinkedIn post about tungsten supply chains.
4:32
SPEAKER_01: But if you can't source the alloys for your Christmas trees,
4:35
SPEAKER_01: your deepwater development doesn't happen.
4:37
SPEAKER_01: Doesn't matter how much you bid on the lease.
4:38
SPEAKER_00: Here's where I push back, though.
4:40
SPEAKER_00: The NPC's been doing studies and issuing reports for decades.
4:43
SPEAKER_00: They're good at framing the problem.
4:45
SPEAKER_00: They're less good at telling you what an individual operator should actually do about it.
4:48
SPEAKER_01: So what?
4:48
SPEAKER_01: You just ignore it?
4:50
SPEAKER_00: No, but the report lands on a shelf at the DOE.
4:52
SPEAKER_00: Maybe it shapes some policy language.
4:54
SPEAKER_00: The operator in the Gulf who just won that block,
4:56
SPEAKER_00: he's not reading an NPC critical mineral study.
4:59
SPEAKER_00: He's calling his procurement team and asking if they can get the valves they need in 18 months.
5:03
SPEAKER_01: And the reason he can't get those valves in 18 months is
5:05
SPEAKER_01: because nobody upstream paid attention to the NPC report five years ago.
5:09
SPEAKER_00: Maybe.
5:10
SPEAKER_00: Or maybe the problem is structural and no report was going to fix it.
5:12
SPEAKER_01: The supply chain problem doesn't announce itself.
5:15
SPEAKER_01: It just shows up as a 12-month lead time on equipment that used to take four.
5:18
SPEAKER_00: I've lived that.
5:20
SPEAKER_00: We had a project in the back end where we couldn't get casing for six weeks
5:23
SPEAKER_00: because every mill was backlogged.
5:25
SPEAKER_00: Nobody saw it coming because nobody was tracking the inputs three levels back.
5:29
SPEAKER_01: So here's a place where automation could actually help.
5:31
SPEAKER_01: Not on the mineral extraction side, that's geopolitics, but on the visibility side.
5:37
SPEAKER_01: If you could map your equipment supply chain back to the raw material source
5:39
SPEAKER_01: and flag concentration risk before it hits your schedule.
5:42
SPEAKER_00: You're talking about visibility across how many vendors, how many tiers?
5:46
SPEAKER_01: Three, maybe four tiers deep.
5:47
SPEAKER_00: And you think those tier three suppliers are going to hand over their sourcing data?
5:51
SPEAKER_01: Not voluntarily.
5:53
SPEAKER_01: But you can infer a lot from trade data, shipping manifests, public filings.
5:57
SPEAKER_01: You don't need every supplier to cooperate.
5:59
SPEAKER_01: You need enough signal to flag when 80% of your nickel alloy is coming
6:02
SPEAKER_01: from one region that just got hit with new export controls.
6:05
SPEAKER_00: Who maintains that, though?
6:06
SPEAKER_00: Who's updating the model when a supplier switches mills or a trade route changes?
6:10
SPEAKER_00: I've seen these supply chain dashboards.
6:12
SPEAKER_00: They look beautiful on day one.
6:14
SPEAKER_00: Six months later,
6:15
SPEAKER_00: the data's stale and the procurement guy's back on the phone doing it the old way.
6:19
SPEAKER_01: You're describing the implementation problem, not the concept problem.
6:22
SPEAKER_01: And you're right that if nobody owns the data pipeline, it rots.
6:25
SPEAKER_01: But the alternative is what you just described.
6:27
SPEAKER_01: Finding out you can't get casing when you're ready to drill.
6:30
SPEAKER_01: That's not a system.
6:31
SPEAKER_01: That's hope with a purchase order attached.
6:33
SPEAKER_00: I didn't say it was a system.
6:34
SPEAKER_00: I said it's what actually happens.
6:36
SPEAKER_01: So the practical answer is probably scoping it tight.
6:39
SPEAKER_01: Don't try to map the whole supply chain for every piece of equipment.
6:42
SPEAKER_01: Pick your 10 longest lead time items.
6:44
SPEAKER_01: Your subsea trees, your high alloy tubulars, whatever is most exposed.
6:47
SPEAKER_01: And build the visibility there first.
6:49
SPEAKER_01: One focused feed, not a platform.
6:51
SPEAKER_00: Ten items, not ten thousand.
6:53
SPEAKER_00: I can see the value in that.
6:54
SPEAKER_01: And then you expand once you've actually proven the data stays current.
6:57
SPEAKER_00: The EOG story's worth mentioning, too.
7:00
SPEAKER_00: They're apparently looking at drilling Eagle Ford-style wells in the UAE,
7:03
SPEAKER_00: which that's a sentence I didn't expect to say.
7:05
SPEAKER_01: EOG in the UAE?
7:07
SPEAKER_00: That's what's being discussed.
7:08
SPEAKER_00: And look, EOG's always been an outlier.
7:11
SPEAKER_00: They've got their own proprietary completions approach, their own rock evaluation methods.
7:15
SPEAKER_00: But exporting that playbook to a completely different geological
7:17
SPEAKER_00: and regulatory environment, that's ambitious.
7:20
SPEAKER_01: More than ambitious.
7:22
SPEAKER_01: The subsurface might be analogous, but the surface game is entirely different.
7:25
SPEAKER_01: You're talking about a national oil company culture,
7:28
SPEAKER_01: different fiscal terms,
7:29
SPEAKER_01: different labor markets, different everything above the drill bit.
7:32
SPEAKER_00: And different reservoir data standards.
7:34
SPEAKER_00: EOG's whole advantage is their data infrastructure,
7:36
SPEAKER_00: the way they evaluate rock quality, the way they optimize spacing.
7:40
SPEAKER_00: That works because they've built it over decades in U.S. shale.
7:43
SPEAKER_00: You drop that into a basin where the data history looks completely different,
7:47
SPEAKER_00: you're not plug and play.
7:48
SPEAKER_01: But that might be precisely why the UAE wants them.
7:50
SPEAKER_01: They've got unconventional resource potential but not the operational DNA to develop it.
7:54
SPEAKER_01: They need someone who's already made the 10,000 mistakes you have to make
7:57
SPEAKER_01: in horizontal development.
7:58
SPEAKER_00: EOG's not a service company, though.
8:00
SPEAKER_00: They're not in the business of teaching other people how to drill.
8:03
SPEAKER_00: So what's the structure?
8:04
SPEAKER_00: JV?
8:04
SPEAKER_00: PSA?
8:05
SPEAKER_00: Are they operating or are they advising?
8:07
SPEAKER_01: The answer probably determines whether this is a real story or just a headline.
8:11
SPEAKER_00: I've seen operators announce international expansions before.
8:13
SPEAKER_00: Half of them quietly shut down the office two years later and take the write-down.
8:17
SPEAKER_01: The other half end up with a position that changes the company's trajectory for 20 years.
8:21
SPEAKER_00: So it's a coin flip with a lot of zeros attached.
8:23
SPEAKER_01: Which is basically the definition of E&P.
8:25
SPEAKER_01: Speaking of compute demand reshaping things,
8:27
SPEAKER_01: there's this conversation happening right now about hidden capacity in
8:30
SPEAKER_01: the existing grid.
8:31
SPEAKER_01: Ramez Nam's been talking about this.
8:34
SPEAKER_01: The claim is there's something like 200 gigawatts just sitting there,
8:37
SPEAKER_01: underutilized in the current U.S. grid infrastructure.
8:39
SPEAKER_00: 200 gigawatts of what, though?
8:41
SPEAKER_01: Transmission and generation capacity that's technically available
8:44
SPEAKER_01: but not being dispatched efficiently.
8:46
SPEAKER_01: The argument is that with better software,
8:48
SPEAKER_01: better load balancing,
8:49
SPEAKER_01: dynamic pricing,
8:50
SPEAKER_01: smarter routing,
8:51
SPEAKER_01: you could unlock a massive chunk of power without building a single new plant.
8:55
SPEAKER_00: So the pitch is we don't need Amazon building a 7.6 gigawatt gas grid in the Permian
8:58
SPEAKER_00: because the power's already there, we're just too dumb to use it.
9:01
SPEAKER_01: The optimistic version.
9:03
SPEAKER_01: And look, there's something to it.
9:05
SPEAKER_01: Grid utilization in the U.S. averages somewhere around 40-50% depending on the region.
9:10
SPEAKER_01: A lot of headroom on paper.
9:11
SPEAKER_00: On paper.
9:12
SPEAKER_00: You're describing a transmission constraint, not a generation constraint.
9:14
SPEAKER_00: Nam's argument focuses more on the generation side,
9:16
SPEAKER_00: but the last mile, or last 12 miles, is where everything breaks down in practice.
9:42
SPEAKER_01: The more interesting thing to me is what this means for AI applied to the grid itself.
9:46
SPEAKER_01: If you can use software to optimize dispatch in real time,
9:48
SPEAKER_01: matching load to generation dynamically instead of
9:50
SPEAKER_01: on these ancient day-ahead scheduling models,
9:52
SPEAKER_01: that could actually move the needle on energy availability.
9:55
SPEAKER_01: Not a chatbot.
9:56
SPEAKER_01: Not a dashboard.
9:57
SPEAKER_01: Actual infrastructure optimization.
9:59
SPEAKER_00: Okay, but who's running that software?
10:01
SPEAKER_00: The ISOs?
10:02
SPEAKER_00: The utilities?
10:03
SPEAKER_00: Because I've dealt with utilities in Texas and Oklahoma,
10:05
SPEAKER_00: and their idea of a software upgrade is migrating from one legacy system to
10:08
SPEAKER_00: a slightly newer legacy system.
10:10
SPEAKER_01: A lot of theoretically in that sentence.
10:11
SPEAKER_01: Granted.
10:32
SPEAKER_01: But the technical concept they're working with,
10:34
SPEAKER_01: this X402 payment rail,
10:36
SPEAKER_01: which is basically using the HTTP 402 status code as a micropayment mechanism
10:41
SPEAKER_01: between AI agents,
10:42
SPEAKER_01: that's the plumbing that would make machine-to-machine transactions possible
10:45
SPEAKER_01: at scale.
10:46
SPEAKER_00: Hold on, you're telling me they want AI systems to pay each other using an HTTP error code?
10:50
SPEAKER_01: Not technically an error code.
10:52
SPEAKER_01: The 402 status was reserved decades ago for this exact purpose,
10:55
SPEAKER_01: payment required, and nobody ever built the infrastructure to use it.
10:59
SPEAKER_01: Now with AI agents that need to transact autonomously, it suddenly makes sense.
11:03
SPEAKER_00: I'm trying to imagine explaining to a controller at a mid-cap operator
11:06
SPEAKER_00: that their power costs are being negotiated in real time by an AI agent using
11:10
SPEAKER_00: a web protocol from the 90s.
11:11
SPEAKER_01: You wouldn't explain it to the controller.
11:13
SPEAKER_01: You'd explain it to the controller's replacement who is also an AI agent.
11:17
SPEAKER_01: Which, okay, I hear myself.
11:19
SPEAKER_01: That's three layers of automation before anyone's checked whether the meter data is accurate.
11:23
SPEAKER_00: Three layers of automation and the first question is still whether
11:25
SPEAKER_00: the meter's right, every time.
11:27
SPEAKER_01: But the Cloudflare play is less about energy and more about the fact
11:30
SPEAKER_01: that AI agents are becoming economic actors.
11:32
SPEAKER_01: They're not just tools anymore.
11:34
SPEAKER_01: They're participants.
11:35
SPEAKER_01: And that changes how you think about everything from data access
11:37
SPEAKER_01: to mineral rights lookups to procurement.
11:39
SPEAKER_00: Let's separate two things.
11:41
SPEAKER_00: There's AI agents doing useful,
11:42
SPEAKER_00: narrow work, optimizing a drilling schedule or flagging title defects.
11:46
SPEAKER_00: And then there's AI agents as autonomous economic participants negotiating with other AI agents.
11:50
SPEAKER_00: Those aren't the same conversation, and I think the second one's getting way ahead of reality.
11:54
SPEAKER_01: Why?
11:55
SPEAKER_01: Genuinely,
11:56
SPEAKER_01: why is agent-to-agent interaction ahead of reality
11:58
SPEAKER_01: when we already have algorithms trading billions of dollars of securities
12:01
SPEAKER_01: against each other every millisecond?
12:03
SPEAKER_00: Because high-frequency trading happens in a market that was designed
12:06
SPEAKER_00: for electronic execution with standardized instruments and clearinghouses.
12:10
SPEAKER_00: The energy grid is a patchwork of utility territories,
12:12
SPEAKER_00: interconnection agreements,
12:14
SPEAKER_00: state-level regulation, and physical infrastructure that was built in the 50s.
12:17
SPEAKER_00: You can't just drop autonomous agents into that and expect them
12:20
SPEAKER_00: to negotiate like it's the NASDAQ.
12:22
SPEAKER_01: The regulatory surface area is completely different.
12:24
SPEAKER_01: I'll concede that.
12:25
SPEAKER_00: And the liability question.
12:27
SPEAKER_00: When an AI agent at a data center negotiates a power draw that causes a brownout in
12:31
SPEAKER_00: a residential area, who's responsible?
12:33
SPEAKER_00: The data center operator?
12:34
SPEAKER_00: The software vendor?
12:36
SPEAKER_00: The agent's, what, parent company?
12:38
SPEAKER_01: So your objection isn't that the technology can't work.
12:40
SPEAKER_01: Your objection is that the institutional and legal framework doesn't exist yet.
12:44
SPEAKER_00: My objection is that people keep building the technology and assuming
12:46
SPEAKER_00: the framework will catch up.
12:48
SPEAKER_00: Sometimes it does.
12:49
SPEAKER_00: Sometimes you end up with a bunch of stranded capability that nobody's allowed to use.
12:53
SPEAKER_01: Which is an argument for building the framework now,
12:54
SPEAKER_01: not for waiting until someone forces us to.
12:57
SPEAKER_00: No, it's an argument for not pretending the technology is further along than
13:00
SPEAKER_00: the world it has to operate in.
13:02
SPEAKER_00: Allie Miller was on a podcast talking about managing a workforce of 34 AI agents
13:06
SPEAKER_00: with an AI chief of staff.
13:07
SPEAKER_00: I'm sure it makes for a great presentation.
13:10
SPEAKER_01: She's building a framework for thinking about AI native organizations.
13:13
SPEAKER_01: The value is conceptual, pointing at where org design might go.
13:16
SPEAKER_00: Everything's conceptual until someone has to run payroll with it.
13:19
SPEAKER_01: Here's what I think is real and worth paying attention to, though.
13:21
SPEAKER_01: Zapier's CMO.
13:23
SPEAKER_01: Dan Slagan, he's got this title, Chief Marketing and AI Transformation Officer.
13:28
SPEAKER_00: Quite a title.
13:29
SPEAKER_01: What he's describing isn't conceptual.
13:31
SPEAKER_01: They're actually rebuilding their internal workflows so that AI isn't a tool you bolt on.
13:35
SPEAKER_01: Every process starts with, can an agent do this?
13:37
SPEAKER_01: And only escalates to a human when it can't.
13:39
SPEAKER_00: Which sounds great until the agent decides it can do something it actually can't,
13:43
SPEAKER_00: and nobody catches it because the human was removed from the loop three steps ago.
13:47
SPEAKER_01: That connects to the OpenAI research that came out.
13:49
SPEAKER_01: Their agents demonstrated the ability to exploit vulnerabilities in
13:52
SPEAKER_01: a simulated hugging-face environment.
13:54
SPEAKER_01: Not maliciously.
13:55
SPEAKER_01: A controlled test.
13:57
SPEAKER_01: But the agents found and used security gaps that
13:59
SPEAKER_01: the researchers didn't explicitly tell them to look for.
14:01
SPEAKER_00: I'm gonna need you to explain why that's not terrifying.
14:04
SPEAKER_01: Because it was a simulation.
14:05
SPEAKER_01: A sandboxed environment specifically designed to test what agents would do
14:08
SPEAKER_01: with elevated permissions.
14:09
SPEAKER_01: The finding isn't that agents are dangerous.
14:11
SPEAKER_01: The finding is that agents are resource-seeking.
14:14
SPEAKER_01: Give them a goal and enough access, and they'll find pathways you didn't anticipate.
14:17
SPEAKER_00: Which in the real world,
14:19
SPEAKER_00: when you're talking about agents managing power procurement or drilling schedules
14:22
SPEAKER_00: or anything with physical consequences,
14:24
SPEAKER_01: means you need to scope their permissions incredibly tightly.
14:26
SPEAKER_01: The same lesson we learned with every junior engineer who ever had admin access to
14:29
SPEAKER_01: a production database.
14:31
SPEAKER_01: The technology isn't the risk.
14:33
SPEAKER_01: The permissions architecture is.
14:34
SPEAKER_00: I've seen what happens when
14:35
SPEAKER_00: a 24-year-old petroleum engineer gets unsupervised access to a completions database.
14:40
SPEAKER_00: So I take your point about permissions.
14:42
SPEAKER_01: And the practical takeaway from all of this,
14:44
SPEAKER_01: whether we're talking about grid optimization,
14:46
SPEAKER_01: agent-to-agent commerce, or AI native workflows, the constraint isn't the model.
14:50
SPEAKER_01: It's governance.
14:51
SPEAKER_01: Who has access to what?
14:53
SPEAKER_01: What's the blast radius when something goes wrong?
14:55
SPEAKER_01: The companies that figure that out first are the ones who will actually deploy at scale.
14:58
SPEAKER_01: Instead of having a beautiful demo, they can't let anyone touch in production.
15:01
SPEAKER_00: Which is 90% of the demos I've seen in the last two years.
15:04
SPEAKER_01: I built three of those demos.
15:05
SPEAKER_01: I said beautiful.
15:07
SPEAKER_01: I didn't say deployable.
15:08
SPEAKER_01: The gap between the two is where most of my humility comes from.
15:11
SPEAKER_00: So here's where I keep landing with all this.
15:13
SPEAKER_00: You've got Amazon dropping a private gas grid into the Permian,
15:16
SPEAKER_00: you've got operators already fighting over pipeline capacity and crews,
15:19
SPEAKER_00: and the one thing nobody seems to have built is a way to see those collisions coming
15:22
SPEAKER_00: before they show up in a landman's title chain.
15:24
SPEAKER_01: Because right now,
15:25
SPEAKER_01: the way an operator finds out that a tech company's infrastructure project is going
15:28
SPEAKER_01: to interfere with their development plan is when someone physically goes
15:31
SPEAKER_01: to run title and discovers a new surface lease,
15:33
SPEAKER_01: or a new easement, or a pipeline right-of-way that wasn't there six months ago.
15:37
SPEAKER_01: And by then you're already behind.
15:39
SPEAKER_00: Or worse, you've already permitted.
15:40
SPEAKER_00: You've got your pad site surveyed,
15:42
SPEAKER_00: you've got your surface use agreement negotiated,
15:44
SPEAKER_00: and then you find out there's a high-voltage transmission corridor coming through
15:47
SPEAKER_00: that quarter section because somebody's feeding a data center 40 miles east.
15:50
SPEAKER_01: So, the question is, can you build something that monitors the signals upstream of that moment?
15:55
SPEAKER_01: And I think you can.
15:56
SPEAKER_01: Not perfectly, but meaningfully.
15:58
SPEAKER_01: You've got Public Utility Commission filings,
16:00
SPEAKER_01: county permit applications,
16:02
SPEAKER_01: pipeline tariff filings at FERSI, satellite imagery showing site prep.
16:06
SPEAKER_01: All of that data exists.
16:07
SPEAKER_01: It's just scattered across 30 different sources,
16:09
SPEAKER_01: and nobody's correlating it against an operator's active development areas.
16:12
SPEAKER_00: The key phrase is active development areas.
16:15
SPEAKER_00: The value isn't in knowing that Amazon's building something in West Texas.
16:18
SPEAKER_00: Everyone knows that.
16:19
SPEAKER_00: The value is in knowing that their transmission easement is going
16:22
SPEAKER_00: to cross your unit 47 to 3H lateral spacing before you've drilled it.
16:25
SPEAKER_01: Spatial overlap detection.
16:27
SPEAKER_01: Take an operator's planned well paths and surface locations,
16:30
SPEAKER_01: overlay them against incoming infrastructure filings and permits
16:32
SPEAKER_01: from non-ENP entities,
16:34
SPEAKER_01: and flag conflicts at the planning stage instead of the execution stage.
16:37
SPEAKER_00: Who inside the operator actually needs that alert, though?
16:40
SPEAKER_00: Because the person running title isn't talking to the person planning well spacing,
16:43
SPEAKER_00: and neither of them is talking to whoever's tracking utility encroachments.
16:47
SPEAKER_00: Those are three different departments in most shops.
16:49
SPEAKER_01: Which is why the output can't just be a dashboard that one team looks at.
16:52
SPEAKER_01: It has to be a structured alert that routes to the right person based on the type of conflict.
16:56
SPEAKER_01: Surface conflict goes to the land team.
16:58
SPEAKER_01: Subsurface spacing issue goes to the development engineer.
17:01
SPEAKER_01: Pipeline capacity competition goes to midstream coordination.
17:04
SPEAKER_00: And then it gets complicated because every operator's org chart is different.
17:07
SPEAKER_00: Some shops have a dedicated right of way group.
17:09
SPEAKER_00: Some shops dump everything on the landman.
17:12
SPEAKER_00: I worked at a place where the division order analyst was somehow responsible
17:15
SPEAKER_00: for tracking easement conflicts because she was the only one who read
17:17
SPEAKER_00: the recorded documents.
17:18
SPEAKER_01: So the system needs configurability at the routing level.
17:21
SPEAKER_01: No, you are not.
17:22
SPEAKER_01: This isn't a real-time collision warning.
17:45
SPEAKER_01: More like a weekly or bi-weekly scan that says,
17:47
SPEAKER_01: here are new filings in your area of interest that might affect your plans.
17:51
SPEAKER_01: And the human still has to validate it.
17:52
SPEAKER_01: You're not auto-rejecting well permits based on an algorithm.
17:55
SPEAKER_00: I'd push back there, though.
17:57
SPEAKER_00: If you're only catching things on a bi-weekly cycle, you might still be too late.
18:00
SPEAKER_00: Some of these infrastructure projects move fast once they're permitted.
18:03
SPEAKER_00: I've seen a solar farm go from filing to grading in six weeks.
18:07
SPEAKER_00: If your scan catches it at week three and the land team doesn't act until week five,
18:10
SPEAKER_00: you're negotiating from a weaker position.
18:12
SPEAKER_01: Counter to that.
18:14
SPEAKER_01: Even a weaker negotiating position is better than no negotiating position,
18:17
SPEAKER_01: which is what you have now when you find out at the title run.
18:20
SPEAKER_01: But I hear the urgency argument.
18:22
SPEAKER_01: So you prioritize the sources by lead time.
18:24
SPEAKER_01: Federal filings, VRC, BLM, right-of-way applications, those tend to have longer timelines.
18:29
SPEAKER_01: You can scan those less frequently.
18:31
SPEAKER_01: County-level permits, especially in active basins, you need to hit more often.
18:34
SPEAKER_00: What about the satellite piece?
18:36
SPEAKER_00: Because that's something a landman can actually act on.
18:39
SPEAKER_00: You see fresh pad construction where there's not supposed to be any.
18:41
SPEAKER_00: You start asking questions.
18:43
SPEAKER_00: That's concrete.
18:44
SPEAKER_01: And verifiable.
18:45
SPEAKER_01: You're not asking someone to trust a prediction.
18:47
SPEAKER_01: You're showing them an image and saying, this changed between May and June.
18:51
SPEAKER_01: Do you know why?
18:52
SPEAKER_01: Fundamentally different conversation than our algorithm thinks there might be a conflict.
18:56
SPEAKER_00: The approval piece is interesting too.
18:58
SPEAKER_00: In most operators I've worked with,
18:59
SPEAKER_00: anything that changes the development plan has to go through some kind of review.
19:03
SPEAKER_00: You can't just move a well location because a satellite image showed a new road.
19:06
SPEAKER_00: There's an engineering review, there's a land review, there's probably an AFE revision.
19:10
SPEAKER_00: So where does this system sit in that workflow?
19:12
SPEAKER_01: Information only, and I feel strongly about that.
19:15
SPEAKER_01: The minute you try to have it recommend moving a well,
19:17
SPEAKER_01: you've crossed into a liability zone that no operator's legal team will accept.
19:21
SPEAKER_01: The system says,
19:23
SPEAKER_01: here's a new permit filing that spatially overlaps with your planned lateral
19:26
SPEAKER_01: in Section 14.
19:27
SPEAKER_01: The development engineer looks at it,
19:29
SPEAKER_01: talks to the land team, and they decide together whether to adjust.
19:32
SPEAKER_01: The system doesn't decide.
19:33
SPEAKER_01: It accelerates the conversation.
19:35
SPEAKER_00: Because the response to a conflict isn't always go fight it.
19:38
SPEAKER_00: Sometimes the right move is to accelerate your own lease negotiation.
19:41
SPEAKER_00: Sometimes you adjust your well path by 200 feet and avoid the whole thing.
19:45
SPEAKER_00: Sometimes you pick up the phone and call the other party's land department
19:48
SPEAKER_00: and work out a shared easement.
19:50
SPEAKER_00: Those are judgment calls.
19:51
SPEAKER_01: The system can't make them.
19:53
SPEAKER_01: But it can make sure the person who does make them has six weeks
19:55
SPEAKER_01: of lead time instead of finding out when the bulldozers show up.
19:57
SPEAKER_00: Here's what I'd add to the build list.
19:59
SPEAKER_00: The output shouldn't just be an alert.
20:01
SPEAKER_00: It should include the filing itself,
20:03
SPEAKER_00: the actual document, plus a map overlay showing the conflict geometry.
20:07
SPEAKER_00: Because when that senior landman walks into the asset team meeting, they need to show, not tell.
20:12
SPEAKER_00: Here's the transmission line route.
20:13
SPEAKER_00: Here's our lateral.
20:14
SPEAKER_00: Here's where they cross.
20:15
SPEAKER_01: So we're building a spatial early warning system with document retrieval and
20:18
SPEAKER_01: a visual conflict layer.
20:19
SPEAKER_01: The hard part is the data ingestion from dozens of different county
20:22
SPEAKER_01: and state filing formats and keeping the operator's footprint current.
20:25
SPEAKER_01: The architecture itself is clear.
20:27
SPEAKER_00: Clear is what every vendor says before they spend 18 months on the county data ingestion piece.
20:31
SPEAKER_01: I didn't say easy.
20:33
SPEAKER_01: There's a gap between a clear architecture and a clean build.
20:35
SPEAKER_01: The county filing formats alone, everyone is different.
20:38
SPEAKER_01: Half are scanned PDFs.
20:40
SPEAKER_01: Some are still on paper in a clerk's drawer.
20:42
SPEAKER_01: That ingestion layer would eat most of the development timeline.
20:45
SPEAKER_00: And nobody's going to fund the glamorous parts until that ugly plumbing works.
20:48
SPEAKER_01: The value here is enormous, though.
20:50
SPEAKER_01: One caught conflict on a single well pad in the Permian could save an operator weeks
20:54
SPEAKER_01: of delay and hundreds of thousands in rerouting costs,
20:56
SPEAKER_01: let alone the legal exposure if you've got competing surface agreements.
20:59
SPEAKER_00: And the operator who would have benefited most
21:01
SPEAKER_00: from this five years ago probably didn't need it because the only things competing
21:04
SPEAKER_00: for surface were other E&P companies and maybe a rancher's fence line.
21:08
SPEAKER_00: Now you've got entities dropping utility scale infrastructure
21:11
SPEAKER_00: that would have been unimaginable in that basin a decade ago.
21:14
SPEAKER_01: 7.6 gigawatts of unimaginable.
21:16
SPEAKER_00: So if we're pulling this all together,
21:18
SPEAKER_00: the build that actually matters this week for me is
21:20
SPEAKER_00: that collision monitoring system, not the concept of it.
21:23
SPEAKER_00: The specific version that watches for infrastructure filings that could interfere
21:27
SPEAKER_00: with an operator's development plan before the operator finds out the hard way.
21:30
SPEAKER_01: And I want to be precise about who this is for, because this isn't a general awareness tool.
21:34
SPEAKER_01: The user is the land team managing undeveloped acreage in a basin
21:37
SPEAKER_01: where non-ENP infrastructure is scaling fast.
21:40
SPEAKER_01: Permian's the obvious one,
21:41
SPEAKER_01: but parts of the Haynesville,
21:42
SPEAKER_01: parts of the Eagle Ford where solar's creeping in, same problem.
21:46
SPEAKER_00: The decision it supports is whether to accelerate a surface negotiation,
21:49
SPEAKER_00: reroute a well path, or escalate to legal before a conflict becomes a fait accompli.
21:54
SPEAKER_00: By the time most land teams find out about a competing surface use,
21:57
SPEAKER_00: the other parties already got a signed agreement and a survey crew on the ground.
22:01
SPEAKER_01: The systems ingesting public filings,
22:02
SPEAKER_01: county permits,
22:03
SPEAKER_01: state utility commission applications,
22:05
SPEAKER_01: railroad commission filings,
22:07
SPEAKER_01: BLM notices if you're on federal,
22:08
SPEAKER_01: and matching those against the operator's plan development footprint.
22:12
SPEAKER_01: Not a two-mile buffer.
22:13
SPEAKER_01: The actual laterals.
22:15
SPEAKER_01: The actual pad locations.
22:16
SPEAKER_01: The access roads.
22:17
SPEAKER_01: The pipeline routes and the development plan.
22:19
SPEAKER_00: And the waiting by project type you mentioned earlier is critical.
22:22
SPEAKER_00: The volume of county permits in a place like Midland or Reeves is enormous.
22:26
SPEAKER_00: Most of it's irrelevant.
22:27
SPEAKER_00: The system needs to understand that a transmission line right-of-way application
22:30
SPEAKER_00: within a quarter mile of your planned access road is urgent,
22:33
SPEAKER_00: and a residential addition three sections over is noise.
22:35
SPEAKER_01: The evidence base is there.
22:37
SPEAKER_01: We've got public filing databases,
22:39
SPEAKER_01: GIS data for planned infrastructure, the operator's own development schedule.
22:43
SPEAKER_01: The gap is that nobody's stitching it together in a way that produces a prioritized,
22:46
SPEAKER_01: actionable alert before the LAN team's morning meeting.
22:48
SPEAKER_00: I want to push back on one thing.
22:50
SPEAKER_00: You said the operator's development schedule.
22:53
SPEAKER_00: In my experience, those schedules change every quarter, sometimes every month.
22:57
SPEAKER_00: The spacing assumptions shift, the timing moves, wells get dropped or added.
23:01
SPEAKER_00: If the system's working off a stale development plan,
23:03
SPEAKER_00: you're back to false positives and missed conflicts.
23:05
SPEAKER_01: So the system needs a lightweight way for the planning team to update the footprint.
23:09
SPEAKER_01: Not a full reimport of their reservoir model.
23:11
SPEAKER_01: Just the surface locations, the lateral azimuths, and the rough timing windows.
23:16
SPEAKER_01: If that update cadence slips, the system should flag that too.
23:19
SPEAKER_01: Hey, your development plan hasn't been refreshed in 90 days
23:22
SPEAKER_01: and there are three new filings in your area.
23:24
SPEAKER_00: I like that self-check.
23:25
SPEAKER_00: I've seen plenty of dashboards that quietly go stale and nobody notices until it's too late.
23:29
SPEAKER_01: The human approval boundary is clear on this one.
23:31
SPEAKER_01: The system identifies and prioritizes potential conflicts.
23:34
SPEAKER_01: It does not initiate contact with the competing party,
23:36
SPEAKER_01: doesn't file anything, doesn't modify the development plan.
23:39
SPEAKER_01: It surfaces a ranked list of potential issues to a human,
23:42
SPEAKER_01: probably the senior landman or the land manager,
23:44
SPEAKER_01: and that person decides what to do with it.
23:46
SPEAKER_00: Because the response varies by situation,
23:48
SPEAKER_00: sometimes you accelerate your lease negotiation,
23:51
SPEAKER_00: sometimes you adjust the well path,
23:52
SPEAKER_00: sometimes you call the other party's land department and work out a shared easement.
23:56
SPEAKER_00: The system can't know which of those is the right move,
23:58
SPEAKER_00: but it can make sure the person with that judgment has six weeks instead
24:01
SPEAKER_00: of six days.
24:02
SPEAKER_01: And the ROI case almost makes itself.
24:05
SPEAKER_01: One avoided mobilization delay pays for years of monitoring.
24:08
SPEAKER_01: You move a rig, you've already got pipe on location, then you discover an easement problem.
24:13
SPEAKER_01: That's a seven figure delay depending on the well program.
24:16
SPEAKER_00: The failure mode I worry about is false positives.
24:18
SPEAKER_00: You start flagging every new county permit within two miles of a well path,
24:22
SPEAKER_00: and the land team ignores the alerts within a month.
24:25
SPEAKER_00: Same problem with any monitoring system.
24:27
SPEAKER_00: If the signal-to-noise ratio is bad, people stop looking.
24:30
SPEAKER_01: Which is why the spatial filtering has to be tight.
24:32
SPEAKER_01: Not within two miles.
24:34
SPEAKER_01: Within the actual footprint of the planned surface disturbance,
24:36
SPEAKER_01: the lateral path, and the access roads.
24:39
SPEAKER_01: And you wait by project type.
24:40
SPEAKER_01: A residential building permit?
24:42
SPEAKER_01: Low priority.
24:43
SPEAKER_01: A utility-scale transmission line filing?
24:46
SPEAKER_01: High priority.
24:47
SPEAKER_01: A new pipeline right-of-way application?
24:49
SPEAKER_01: Immediate.
24:50
SPEAKER_00: So you're asking the system to understand the difference between a barn and a substation?
24:54
SPEAKER_01: A barn with seven nuclear reactors' worth of power?
24:56
SPEAKER_01: Demand behind it, sure.
24:57
SPEAKER_01: Which brings us full circle to where we started.
25:00
SPEAKER_01: Nobody needed this system when the biggest surface conflict was a stock tank placement.
25:03
SPEAKER_01: The basin changed faster than the tools.
25:05
SPEAKER_00: The governance question, who has access to the operator's development plans inside this system?
25:10
SPEAKER_00: Because that's sensitive data.
25:12
SPEAKER_00: Well, locations, spacing assumptions, timing.
25:14
SPEAKER_00: You don't want that leaking.
25:15
SPEAKER_00: And if you're a service provider running this for multiple operators,
25:18
SPEAKER_00: you've got a Chinese wall problem immediately.
25:20
SPEAKER_01: Probably the biggest constraint on who can actually build this.
25:23
SPEAKER_01: A third-party vendor running it for multiple operators in the same basin?
25:26
SPEAKER_01: The data segregation requirements are enormous.
25:28
SPEAKER_01: You'd almost need to deploy it as an on-premise tool where
25:31
SPEAKER_01: the operator's data never leaves their environment.
25:33
SPEAKER_01: And only the external monitoring,
25:35
SPEAKER_01: the public filings, the satellite feeds, comes in from outside.
25:38
SPEAKER_00: Limits your scale but protects your client.
25:41
SPEAKER_00: Most operators would take that trade-off.
25:43
SPEAKER_01: The operator who could benefit most from this right now is anyone with a large,
25:46
SPEAKER_01: undeveloped acreage position in the Permian.
25:48
SPEAKER_01: The infrastructure collision density is highest there.
25:50
SPEAKER_01: You've got E&P.
25:52
SPEAKER_01: You've got midstream build-out.
25:53
SPEAKER_01: You've got solar.
25:54
SPEAKER_01: You've got data center power infrastructure.
25:56
SPEAKER_01: You've got water recycling facilities.
25:57
SPEAKER_01: All competing for the same surface.
25:59
SPEAKER_00: Five years ago,
26:00
SPEAKER_00: if you told me a landman's biggest headache in the Permian would be
26:02
SPEAKER_00: a tech company's power grid, I probably would have believed you.
26:06
SPEAKER_00: But I wouldn't have liked it.
26:07
SPEAKER_01: The landman who sees it coming six weeks early is the one who keeps the drill schedule intact.
26:11
SPEAKER_01: The whole pitch here isn't AI will solve your problems.
26:14
SPEAKER_01: The pitch is you'll stop being surprised by things that were already public information.
26:18
SPEAKER_00: I can sell that to an asset team.
26:20
SPEAKER_01: And if you're listening to this thinking, I actually need that built for my operation.
26:24
SPEAKER_01: Reach out.
26:25
SPEAKER_01: Siliconandshale at gmail.com.
26:27
SPEAKER_01: We build this stuff.
26:28
SPEAKER_00: And we'll make sure the county filing formats don't make you question your career choices.
26:32
SPEAKER_00: That part's on us.
26:33
SPEAKER_00: What gets me is we spent this whole conversation talking about conflicts
26:36
SPEAKER_00: between operators and tech companies.
26:38
SPEAKER_00: And five years ago, the biggest surface dispute I ever saw was over a stock tank placement.
26:42
SPEAKER_01: And now somebody's dropping a small country's worth of power generation into the same counties.
26:46
SPEAKER_01: The scale compression is genuinely hard to wrap your head around.
26:49
SPEAKER_00: The tools will catch up.
26:50
SPEAKER_00: They always do.
26:52
SPEAKER_00: But the judgment piece,
26:53
SPEAKER_00: the part where somebody's got to look at a map and a lease and a relationship
26:56
SPEAKER_00: and figure out the actual move, that's not going anywhere.
26:59
SPEAKER_01: On that we completely agree.
27:00
SPEAKER_01: Oh, and if your company sells to operators,
27:02
SPEAKER_01: landmen,
27:03
SPEAKER_01: anyone in the energy space,
27:04
SPEAKER_01: and you want to reach the folks who actually make buying decisions,
27:06
SPEAKER_01: we're taking sponsors.
27:08
SPEAKER_01: Siliconandshale at gmail.com.
27:10
SPEAKER_01: We'll use your product before we talk about it.
27:12
SPEAKER_00: We will, and if it doesn't work, we'll talk about that too.
27:14
SPEAKER_01: Probably a better sales pitch than it sounds like.
27:16
SPEAKER_00: Pack,
27:17
SPEAKER_00: I've got a call with a surface owner in 20 minutes who I'm pretty sure doesn't know
27:19
SPEAKER_00: a substation's going in next to his calving pasture.
27:22
SPEAKER_01: Go handle that.
27:23
SPEAKER_01: We'll pick this up next week.
27:24
SPEAKER_00: If the calving pasture's still there.