Latest / Silicon & Shale Podcast / Episode 7: The Grid Crunch Hits Main Street

Transcript

Episode 7: The Grid Crunch Hits Main Street · Silicon & Shale Podcast

• 0:00
  SPEAKER_01: I'm telling you, Derek, the energy crunch isn't just a big city problem.
• 0:03
  SPEAKER_01: Small-town America finds itself suddenly grappling with the demands
• 0:06
  SPEAKER_01: of these massive data centers.
• 0:07
  SPEAKER_01: Happens almost overnight.
• 0:09
  SPEAKER_00: I saw that.
• 0:10
  SPEAKER_00: The Ohio Brickmaker, 141 years old.
• 0:13
  SPEAKER_00: Poof.
• 0:14
  SPEAKER_00: Gone.
• 0:15
  SPEAKER_00: A server farm decided it needed their power.
• 0:17
  SPEAKER_00: That's a hell of a way to close shop.
• 0:19
  SPEAKER_00: Not because your bricks failed, but because some blinky lights needed more juice.
• 0:23
  SPEAKER_01: It's a bizarre trend.
• 0:24
  SPEAKER_01: And it's not isolated to Ohio.
• 0:26
  SPEAKER_01: We're tracking this nationwide, from rural Georgia all the way to parts of Oklahoma.
• 0:31
  SPEAKER_01: These data centers demand immense power, often seeking cheap land and cheap electricity.
• 0:36
  SPEAKER_01: That means they're not always settling in established industrial zones.
• 0:39
  SPEAKER_00: So they just pop up.
• 0:42
  SPEAKER_00: Like mushrooms after a rain,
• 0:43
  SPEAKER_00: only these are giant buildings packed with humming computers,
• 0:46
  SPEAKER_00: sucking enough power to run a small city.
• 0:48
  SPEAKER_00: Then the local co-op has to figure out how to feed them.
• 0:51
  SPEAKER_00: That's a problem, a genuine grid crunch.
• 0:53
  SPEAKER_00: We discussed it last week, didn't we?
• 0:55
  SPEAKER_00: The grid.
• 0:56
  SPEAKER_00: Always the grid.
• 0:57
  SPEAKER_01: It's the exact same issue, only manifesting differently.
• 1:01
  SPEAKER_01: Last week, we discussed AI's broader energy demand straining the grid.
• 1:04
  SPEAKER_01: This is the hyper-local version, directly impacting businesses and communities.
• 1:08
  SPEAKER_01: It makes me wonder, what does this mean for our industry?
• 1:12
  SPEAKER_01: When a new data center suddenly requires the power of a mid-sized gas processing plant?
• 1:16
  SPEAKER_00: It's already doing it, Pac.
• 1:17
  SPEAKER_00: No need to wonder.
• 1:18
  SPEAKER_01: Welcome to Silicon and Shale, where the oil patch meets the algorithm.
• 1:22
  SPEAKER_01: Week of July 20th, 2026.
• 1:25
  SPEAKER_00: I'm Derek.
• 1:26
  SPEAKER_01: And I'm Packer.
• 1:27
  SPEAKER_01: We're tackling that very problem today because it isn't a theoretical future issue.
• 1:32
  SPEAKER_01: It's hitting us right now,
• 1:33
  SPEAKER_01: especially when you consider the economics
• 1:35
  SPEAKER_01: of powering places suddenly demanding huge loads.
• 1:37
  SPEAKER_01: We'll also discuss how AI agents are stepping beyond mere office workflows.
• 1:41
  SPEAKER_01: Pretty incredible.
• 1:43
  SPEAKER_00: Yeah, we'll explore how those agents might actually get out there and do some actual work.
• 1:47
  SPEAKER_00: And I've got a courthouse chronicle coming up that might surprise you,
• 1:50
  SPEAKER_00: considering all this talk about power and infrastructure.
• 1:53
  SPEAKER_01: Oh, I'm ready.
• 1:53
  SPEAKER_01: Let's dig in.
• 1:54
  SPEAKER_00: So, these data centers, all that power they're drawing, that's not solely a domestic issue.
• 2:01
  SPEAKER_00: We're watching geopolitical events that could upend the whole thing.
• 2:04
  SPEAKER_00: Take the Strait of Hormuz, for example.
• 2:06
  SPEAKER_00: I heard talk about a potential 20% toll on oil tankers passing through there.
• 2:11
  SPEAKER_00: 20%.
• 2:11
  SPEAKER_00: You hear that and you just gulp.
• 2:14
  SPEAKER_01: 20% is massive.
• 2:15
  SPEAKER_01: We're discussing a huge portion of global oil transiting that choke point.
• 2:19
  SPEAKER_01: A 20% toll,
• 2:20
  SPEAKER_01: on top of everything else,
• 2:21
  SPEAKER_01: with crude already nearing $90 a barrel, will send shocks through global markets.
• 2:25
  SPEAKER_01: It feels like kicking us when we're down.
• 2:27
  SPEAKER_01: Especially with these healthy truce violations we've seen.
• 2:30
  SPEAKER_00: And you're already navigating a tight market,
• 2:32
  SPEAKER_00: global instability, and then someone proposes something like that.
• 2:36
  SPEAKER_00: It piles on more uncertainty.
• 2:38
  SPEAKER_00: For operators, for anyone in the supply chain, that's a massive wildcard.
• 2:42
  SPEAKER_00: You're trying to plan, trying to hedge, and then you have this potential cost spike lingering.
• 2:47
  SPEAKER_01: And it isn't the direct cost of the oil alone.
• 2:49
  SPEAKER_01: It's the ripple effect on every oil-dependent sector.
• 2:52
  SPEAKER_01: Shipping costs for all goods,
• 2:53
  SPEAKER_01: manufacturing,
• 2:54
  SPEAKER_01: even the price of bricks from that old brickmaker,
• 2:56
  SPEAKER_01: assuming they're still in business.
• 2:59
  SPEAKER_01: This injects extreme volatility into every calculation.
• 3:02
  SPEAKER_01: It's also a stark reminder of how deeply interconnected everything remains.
• 3:06
  SPEAKER_01: We discuss AI and data centers,
• 3:08
  SPEAKER_01: but the physical world, geopolitics, they still hold immense sway.
• 3:12
  SPEAKER_00: You can build the most brilliant AI imaginable,
• 3:15
  SPEAKER_00: but if you can't move crude to the refinery or gas to the power plant,
• 3:18
  SPEAKER_00: well, you just sit there, don't you?
• 3:20
  SPEAKER_00: It comes down to that physical infrastructure, the very flow of things.
• 3:24
  SPEAKER_00: And when that flow chokes, whether by a toll or by actual conflict, everything grinds to a halt.
• 3:29
  SPEAKER_01: Sure, but the best model's only as good as the garbage you feed it.
• 3:31
  SPEAKER_01: When you've got a healthy missile coming down or some major player suddenly changes policy,
• 4:02
  SPEAKER_00: Human
• 4:07
  SPEAKER_01: unpredictability is the wild card.
• 4:09
  SPEAKER_01: But imagine if these models could ingest real-time shipping data,
• 4:12
  SPEAKER_01: satellite imagery,
• 4:13
  SPEAKER_01: geopolitical news feeds,
• 4:15
  SPEAKER_01: and then,
• 4:15
  SPEAKER_01: based on historical patterns of similar incidents,
• 4:18
  SPEAKER_01: give you a range of potential outcomes,
• 4:20
  SPEAKER_01: complete with their probabilities,
• 4:22
  SPEAKER_01: It wouldn't tell you what will happen,
• 4:23
  SPEAKER_01: but it might say,
• 4:24
  SPEAKER_01: there's an 80% chance of a 5% price hike,
• 4:27
  SPEAKER_01: but also a 20% chance of a 15% spike if X, Y, or Z occurs.
• 4:32
  SPEAKER_00: Hmm, that's, that's certainly interesting.
• 4:35
  SPEAKER_00: Right now you've got a guy in a trading room or a landman trying to figure out if
• 4:39
  SPEAKER_00: a deal's still gonna pencil out and what are they doing?
• 4:41
  SPEAKER_00: Mostly watching the news, aren't they?
• 4:43
  SPEAKER_00: Maybe calling a buddy who knows a guy.
• 4:45
  SPEAKER_00: It's all gut feeling based on years of experience.
• 4:48
  SPEAKER_00: And I trust experience,
• 4:49
  SPEAKER_00: but if you could augment that,
• 4:51
  SPEAKER_01: The goal isn't to replace that experience.
• 4:53
  SPEAKER_01: It's to give that seasoned person more data,
• 4:55
  SPEAKER_01: faster, presented in a way that helps them spot patterns they might otherwise miss.
• 5:00
  SPEAKER_01: Jag Gattu from Uptime AI,
• 5:02
  SPEAKER_01: for instance,
• 5:03
  SPEAKER_01: spoke about AI transforming decision-making in upstream operations,
• 5:06
  SPEAKER_01: specifically focusing on data analytics and sensor technology.
• 5:10
  SPEAKER_01: He sees machines helping humans make better decisions, not making them for them.
• 5:14
  SPEAKER_01: That's what being proactive means.
• 5:15
  SPEAKER_00: Proactive.
• 5:17
  SPEAKER_00: Yeah, everyone wants to be ahead of the curve, don't they?
• 5:20
  SPEAKER_00: In this business, a single day or two can mean millions.
• 5:23
  SPEAKER_00: So if your sensors are telling you a pump's about to go down, that's one thing.
• 5:27
  SPEAKER_00: If an AI is telling you the price of crude will jump because
• 5:30
  SPEAKER_00: of some political maneuver halfway across the world,
• 5:32
  SPEAKER_00: now that's a whole other level of proactive.
• 5:35
  SPEAKER_01: And integrating those data streams, that's crucial as well.
• 5:38
  SPEAKER_01: Not just sensor data from the well site,
• 5:40
  SPEAKER_01: but also global economic indicators, geopolitical intelligence.
• 5:44
  SPEAKER_01: That presents both the real challenge and the real opportunity.
• 5:47
  SPEAKER_01: Currently,
• 5:48
  SPEAKER_01: these streams are often siloed You have one team watching the well,
• 5:51
  SPEAKER_01: another tracking markets, another monitoring news.
• 5:54
  SPEAKER_01: What if we could bring it all together?
• 5:56
  SPEAKER_00: Reminds me of a situation back in Grady County, oh, probably 15 years ago now.
• 6:01
  SPEAKER_00: We had a pretty active play going on there.
• 6:03
  SPEAKER_00: A lot of drilling and the county clerk's office was swamped.
• 6:06
  SPEAKER_00: Paper everywhere.
• 6:07
  SPEAKER_00: Deeds, leases, pooling orders.
• 6:09
  SPEAKER_00: Everything was a physical document.
• 6:10
  SPEAKER_00: And the oil price was bouncing around like a rubber ball which made folks real antsy.
• 6:15
  SPEAKER_01: Oh, a Grady County Courthouse Chronicle.
• 6:17
  SPEAKER_01: I love these.
• 6:18
  SPEAKER_00: Yeah, this one's a classic.
• 6:19
  SPEAKER_00: Me and a couple of other landmen in there, wrestling title work on a big unit.
• 6:24
  SPEAKER_00: The clerk's office, they'd just gotten a new fax machine.
• 6:26
  SPEAKER_00: A brand new one, mind you.
• 6:27
  SPEAKER_00: They were beaming.
• 6:28
  SPEAKER_00: Said it would speed everything up.
• 6:30
  SPEAKER_01: The fax machine is still warm.
• 6:32
  SPEAKER_00: That's the one.
• 6:33
  SPEAKER_00: The fax machine is still warm.
• 6:35
  SPEAKER_00: I remember thinking, bless their hearts, they think this is progress.
• 6:39
  SPEAKER_00: Meanwhile,
• 6:39
  SPEAKER_00: we're all still waiting for someone to physically pull a book,
• 6:42
  SPEAKER_00: find a page, make a copy, and then maybe fax it to us.
• 6:45
  SPEAKER_00: The price of oil would swing five bucks a barrel in the time it took to get one document.
• 6:49
  SPEAKER_01: So if they'd had an AI agent or even basic digitization to scan
• 6:53
  SPEAKER_01: and index those documents, you'd think it would have been a game changer?
• 6:57
  SPEAKER_00: A game changer?
• 6:58
  SPEAKER_00: Packer, it would have been a revolution.
• 7:00
  SPEAKER_00: We watched people lose money because they couldn't get information fast enough
• 7:03
  SPEAKER_00: to react to market shifts.
• 7:05
  SPEAKER_00: A new lease, a division order, a change in ownership.
• 7:08
  SPEAKER_00: If you couldn't get that recorded and confirmed quickly, you were flying blind.
• 7:11
  SPEAKER_00: And in a volatile market, flying blind is a good way to crash.
• 7:14
  SPEAKER_01: That's... that's a stark example.
• 7:17
  SPEAKER_01: Information latency, even at the local level, can carry huge financial implications.
• 7:22
  SPEAKER_01: The issue isn't always global macro trends.
• 7:25
  SPEAKER_01: Often it's just waiting for a county clerk to locate a deed book.
• 7:28
  SPEAKER_00: It's the boots on the ground pack,
• 7:30
  SPEAKER_00: always, and what happens when those boots are stuck waiting for a piece of paper.
• 7:34
  SPEAKER_00: So when I hear about these AI reasoning agents,
• 7:36
  SPEAKER_00: I think, okay,
• 7:37
  SPEAKER_00: can they find that deed for me in Grady County and tell me if it's got
• 7:40
  SPEAKER_00: a proper acknowledgement?
• 7:42
  SPEAKER_00: If they can do that, then we're really talking.
• 7:44
  SPEAKER_01: Things get interesting when an AI agent doesn't just find the deed, but understands it.
• 7:49
  SPEAKER_01: Imagine it parsing legal language,
• 7:51
  SPEAKER_01: flagging potential issues, and cross-referencing with other documents in the chain.
• 7:55
  SPEAKER_01: This moves beyond a simple search engine.
• 7:57
  SPEAKER_01: It becomes true expertise augmentation.
• 7:59
  SPEAKER_00: So it could sniff out a bad title chain from three counties away?
• 8:02
  SPEAKER_01: You've got the idea.
• 8:04
  SPEAKER_01: Not three counties,
• 8:05
  SPEAKER_01: perhaps,
• 8:05
  SPEAKER_01: but it could certainly flag inconsistencies within the same county or even inside
• 8:09
  SPEAKER_01: a single document chain.
• 8:11
  SPEAKER_01: These are the details that take a seasoned landman hours, sometimes days to uncover.
• 8:15
  SPEAKER_01: That capability frees up the landman for the truly complex work,
• 8:19
  SPEAKER_01: negotiation, relationship building, the human elements AI can't touch.
• 8:23
  SPEAKER_00: I'll tell you this,
• 8:24
  SPEAKER_00: if it saves me a trip to a courthouse basement with no cell signal,
• 8:27
  SPEAKER_00: I'll call that a win.
• 8:29
  SPEAKER_00: And if it does it accurately, then we might have something real.
• 8:32
  SPEAKER_00: Right now, I'm one bad PDF scan from retirement.
• 8:35
  SPEAKER_01: Put it on the Midnight Build list, a title agent that can survive a bad PDF scan.
• 8:39
  SPEAKER_01: If your shop has a problem like that,
• 8:41
  SPEAKER_01: something everybody gripes about but nobody has time to build,
• 8:44
  SPEAKER_01: email siliconandshale at gmail.com.
• 8:47
  SPEAKER_01: We'll see if there's a practical way to tackle it.
• 8:49
  SPEAKER_00: Superpowers, huh?
• 8:50
  SPEAKER_00: I like the sound of that.
• 8:52
  SPEAKER_00: Provided those superpowers don't involve me learning a new programming language.
• 8:55
  SPEAKER_01: No coding homework, promise.
• 8:57
  SPEAKER_01: Let me throw you a curveball.
• 8:58
  SPEAKER_01: The 2028 LA Olympics.
• 9:00
  SPEAKER_00: The Olympics?
• 9:01
  SPEAKER_00: What's that got to do with oil and gas,
• 9:03
  SPEAKER_00: or AI for that matter, besides the traffic and the crowds?
• 9:06
  SPEAKER_01: Traffic is part of it.
• 9:08
  SPEAKER_01: The crowds, too, and the massive logistical challenge of hosting an event that size.
• 9:12
  SPEAKER_01: But the Energy Gang podcast brought up how
• 9:14
  SPEAKER_01: the LA Olympics might catalyze clean energy initiatives.
• 9:18
  SPEAKER_01: They see it as a unique chance to transform transport, power, and air quality across the city.
• 9:22
  SPEAKER_00: So a giant temporary influx of people is supposed to improve air quality?
• 9:26
  SPEAKER_00: That strikes me as a contradiction.
• 9:28
  SPEAKER_00: Drop a million folks into one place,
• 9:30
  SPEAKER_00: you generally get a million cars and a whole lot of exhaust fumes.
• 9:33
  SPEAKER_01: That's the core idea, though.
• 9:35
  SPEAKER_01: The event's sheer scale compels innovation.
• 9:38
  SPEAKER_01: LA must figure out how to transport those people,
• 9:40
  SPEAKER_01: power those venues,
• 9:41
  SPEAKER_01: and maintain breathable air, all while presenting a green image globally.
• 9:46
  SPEAKER_01: So they're considering massive EV infrastructure build-outs,
• 9:48
  SPEAKER_01: temporary microgrids, and public transit overhauls.
• 9:51
  SPEAKER_01: It's a real-world stress test for clean energy solutions.
• 9:54
  SPEAKER_00: Okay, I track what you're saying.
• 9:56
  SPEAKER_00: It's a huge fixed deadline with many eyes watching.
• 9:59
  SPEAKER_00: So it forces them to actually do things, not just jaw about them.
• 10:02
  SPEAKER_00: But how does that ripple into our E&P world,
• 10:05
  SPEAKER_00: aside from perhaps a temporary dip in gasoline demand around LA for a few weeks?
• 10:09
  SPEAKER_01: It goes deeper than that.
• 10:10
  SPEAKER_01: Consider the infrastructure built.
• 10:12
  SPEAKER_01: The charging stations, the smart grid technologies, the energy storage solutions.
• 10:16
  SPEAKER_01: They don't just serve the Olympics.
• 10:18
  SPEAKER_01: They become permanent installations.
• 10:20
  SPEAKER_01: And the lessons learned,
• 10:21
  SPEAKER_01: the best practices for managing power demand and supply,
• 10:23
  SPEAKER_01: for integrating renewables massively.
• 10:25
  SPEAKER_01: Those are transferable assets.
• 10:27
  SPEAKER_00: So it amounts to a big expensive pilot project for the whole country,
• 10:30
  SPEAKER_00: potentially, or at minimum for other major cities.
• 10:33
  SPEAKER_01: That captures the essence.
• 10:35
  SPEAKER_01: For us, it means observing what works, what doesn't.
• 10:38
  SPEAKER_01: Which demand response programs prove effective?
• 10:41
  SPEAKER_01: How reliable are these new microgrids under extreme stress?
• 10:44
  SPEAKER_01: What are the actual costs and benefits?
• 10:47
  SPEAKER_01: If LA manages this for the Olympics,
• 10:49
  SPEAKER_01: other cities, other industries might adopt these solutions more seriously.
• 10:53
  SPEAKER_01: That could open new markets for natural gas,
• 10:55
  SPEAKER_01: for instance, as a backup for renewables or for hydrogen.
• 10:58
  SPEAKER_00: Now I see the angle.
• 11:00
  SPEAKER_00: Going green is the headline.
• 11:02
  SPEAKER_00: The method tells you what happens to the broader energy mix.
• 11:05
  SPEAKER_00: You won't power the Olympics solely with solar panels and wishful thinking, will you?
• 11:08
  SPEAKER_00: You'll still require reliable, dispatchable power.
• 11:11
  SPEAKER_01: Our industry still has a critical role to play.
• 11:14
  SPEAKER_01: We must understand how our resources fit into this evolving energy landscape
• 11:18
  SPEAKER_01: and how we become part of the solution.
• 11:20
  SPEAKER_01: Agility's key.
• 11:21
  SPEAKER_01: The oil
• 11:22
  SPEAKER_01: and gas industry always showed remarkable agility-solving complex engineering
• 11:25
  SPEAKER_01: challenges.
• 11:26
  SPEAKER_00: Agile, sure.
• 11:27
  SPEAKER_00: We pivot when prices swing.
• 11:29
  SPEAKER_00: We find new ways to pull crude from the earth.
• 11:32
  SPEAKER_00: But adapting to a city trying for carbon neutral for a couple of weeks?
• 11:35
  SPEAKER_00: That's a whole different kind of pivot.
• 11:37
  SPEAKER_00: Drilling deeper won't solve it.
• 11:39
  SPEAKER_00: It requires new thinking.
• 11:40
  SPEAKER_01: This is where AI becomes crucial.
• 11:42
  SPEAKER_01: Optimizing energy distribution,
• 11:44
  SPEAKER_01: predicting demand spikes, managing complex charging networks for thousands of EVs.
• 11:49
  SPEAKER_01: That's a massive data and optimization challenge.
• 11:51
  SPEAKER_01: These are the problems AI agents are built to handle.
• 11:54
  SPEAKER_00: So these AI agents, they could basically run the entire LA power grid for the Olympics?
• 11:59
  SPEAKER_00: Tell me when to fire up the natural gas plant,
• 12:01
  SPEAKER_00: when to shut down the solar, where to send the juice.
• 12:04
  SPEAKER_01: Not necessarily run it autonomously,
• 12:05
  SPEAKER_01: but definitely optimize it,
• 12:07
  SPEAKER_01: provide recommendations,
• 12:08
  SPEAKER_01: predict outages, manage load balancing across distributed resources.
• 12:12
  SPEAKER_01: Picture real-time data from every charging station,
• 12:15
  SPEAKER_01: every solar array,
• 12:16
  SPEAKER_01: every building,
• 12:18
  SPEAKER_01: all feeding into an agent constantly seeking the most efficient,
• 12:20
  SPEAKER_01: reliable demand fulfillment.
• 12:22
  SPEAKER_01: That's a significant leap beyond traditional grid management.
• 12:25
  SPEAKER_00: Those are a lot of variables.
• 12:27
  SPEAKER_00: A lot of variables.
• 12:28
  SPEAKER_00: The human brain, even the best grid operator, juggles only so many things simultaneously.
• 12:33
  SPEAKER_00: But if an AI sees all those pieces at once and makes sense of them,
• 12:37
  SPEAKER_00: yeah, I get the appeal there.
• 12:38
  SPEAKER_00: The scale of that complexity is pretty wild.
• 12:41
  SPEAKER_01: AI earns its keep on problems like this.
• 12:44
  SPEAKER_01: Too many inputs changing too quickly for one operator to hold in their head.
• 12:47
  SPEAKER_01: Real-time decisions, competing objectives, no room for a blackout.
• 12:51
  SPEAKER_01: That's a long way from spreadsheets and marketing.
• 12:53
  SPEAKER_01: It's fundamental infrastructure.
• 12:55
  SPEAKER_00: Well, I'll tell you what.
• 12:57
  SPEAKER_00: If it means fewer blackouts during the pole vault, I'm all for it.
• 13:00
  SPEAKER_00: Nothing kills the Olympic spirit faster than a power outage during the opening ceremony.
• 13:04
  SPEAKER_01: Fewer blackouts would be a start.
• 13:06
  SPEAKER_01: And that's the useful part of these agents.
• 13:08
  SPEAKER_01: They're finally moving beyond theory.
• 13:10
  SPEAKER_00: Doing real work, not just pondering how to do real work.
• 13:13
  SPEAKER_01: Brian Good from Microsoft was discussing concepts for managing
• 13:16
  SPEAKER_01: and deploying AI agents in businesses.
• 13:18
  SPEAKER_01: Think of it as a control plane.
• 13:20
  SPEAKER_01: As these agents become widespread,
• 13:22
  SPEAKER_01: companies need to manage them securely, deploy effectively, and ensure performance.
• 13:26
  SPEAKER_00: So, a boss for the robots?
• 13:28
  SPEAKER_00: Keeping them from wandering off doing their own thing or maybe wiping out
• 13:31
  SPEAKER_00: the company's whole lease database by accident?
• 13:34
  SPEAKER_00: PAC.
• 13:34
  SPEAKER_00: That's my nightmare.
• 13:36
  SPEAKER_00: My absolute nightmare.
• 13:37
  SPEAKER_01: You've got it.
• 13:39
  SPEAKER_01: Governance and security are key.
• 13:41
  SPEAKER_01: If autonomous agents run across your entire enterprise,
• 13:44
  SPEAKER_01: optimizing supply chains,
• 13:45
  SPEAKER_01: handling customer service,
• 13:46
  SPEAKER_01: even analyzing geological data,
• 13:48
  SPEAKER_01: ensure they align with business objectives, avoid vulnerabilities.
• 13:52
  SPEAKER_01: It's the unglamorous but essential piece for scaling AI.
• 13:55
  SPEAKER_00: Boring, perhaps, but I'd say boring keeps the lights on and the lawsuits at bay.
• 13:59
  SPEAKER_00: So, if Microsoft's putting thought into controlling these things,
• 14:02
  SPEAKER_00: that tells me they're serious about deploying them in the real world,
• 14:05
  SPEAKER_00: not just keeping them in some lab.
• 14:07
  SPEAKER_01: John Bickett, Samsara's co-founder and CTO, discussed the physical side at Samsara Beyond 2026.
• 14:13
  SPEAKER_01: He spoke about expanding AI agents beyond typical office workflows
• 14:16
  SPEAKER_01: to benefit fleets, drivers, and infrastructure.
• 14:19
  SPEAKER_00: Now that gets interesting.
• 14:20
  SPEAKER_00: Fleets, drivers, physical infrastructure?
• 14:22
  SPEAKER_00: You mean actual trucks out there on the road?
• 14:25
  SPEAKER_01: Samsara weaves AI right into dashcams, telematics gear, and asset trackers.
• 14:30
  SPEAKER_01: Picture this.
• 14:31
  SPEAKER_01: An AI agent constantly watches a truck driver's behavior through the dashcam,
• 14:35
  SPEAKER_01: spotting fatigue, distracted driving, or even poor route choices.
• 14:39
  SPEAKER_01: Another agent tracks equipment performance in the field,
• 14:41
  SPEAKER_01: predicting maintenance long before a critical failure happens.
• 14:44
  SPEAKER_00: Hold on, the dashcam has an AI inside, watching the driver?
• 14:47
  SPEAKER_00: That feels pretty big brother, doesn't it?
• 14:50
  SPEAKER_00: I get the safety side.
• 14:52
  SPEAKER_00: Still, that's a lot of eyes.
• 14:53
  SPEAKER_01: Privacy is a big part of the discussion, for sure.
• 14:56
  SPEAKER_01: But the point is safety and efficiency.
• 14:59
  SPEAKER_01: If an AI senses a driver drifting or notices they've gone too long without a brake
• 15:03
  SPEAKER_01: and then gives a soft alert, that stops a catastrophic accident.
• 15:06
  SPEAKER_01: For fleet managers, it means route optimization, less fuel burned, longer vehicle life.
• 15:11
  SPEAKER_01: We're talking operational excellence.
• 15:13
  SPEAKER_00: So this isn't simply you're speeding.
• 15:16
  SPEAKER_00: It's more like you've been behind the wheel for eight hours,
• 15:18
  SPEAKER_00: maybe grab a coffee,
• 15:19
  SPEAKER_00: or that engine's sounding off, better get a check before it seizes.
• 15:23
  SPEAKER_00: That's genuinely practical,
• 15:26
  SPEAKER_00: especially for our crews working out there covering those long hauls
• 15:28
  SPEAKER_00: between well sites.
• 15:30
  SPEAKER_01: Think about a completions crew driving from Oklahoma City all the way out to Roger Mills County.
• 15:34
  SPEAKER_01: That's a serious haul.
• 15:36
  SPEAKER_01: If an AI in their truck could help them navigate,
• 15:38
  SPEAKER_01: monitor their driving,
• 15:39
  SPEAKER_01: optimize their stops,
• 15:40
  SPEAKER_01: even warn them about potential road hazards based on real-time data,
• 15:43
  SPEAKER_01: that's a huge benefit.
• 15:45
  SPEAKER_01: Less downtime, safer operations.
• 15:47
  SPEAKER_00: Roger Mills, eh?
• 15:48
  SPEAKER_00: Yeah, that's a long one.
• 15:50
  SPEAKER_00: Just miles and miles of nothing but prairie.
• 15:52
  SPEAKER_00: If your truck crap's out there, you're in a world of hurt.
• 15:55
  SPEAKER_00: So, an AI predicting a breakdown?
• 15:57
  SPEAKER_00: That's actually pretty compelling.
• 15:59
  SPEAKER_00: I'll give that a solid 7 on the fax machine to fully autonomous scale.
• 16:03
  SPEAKER_00: Not fully autonomous yet, but it's definitely way past the fax machine.
• 16:07
  SPEAKER_00: We're talking real-time, preventative.
• 16:09
  SPEAKER_01: Ooh, a 7.
• 16:11
  SPEAKER_01: High praise, Derek.
• 16:11
  SPEAKER_01: I'll take it.
• 16:13
  SPEAKER_01: And this doesn't stop with vehicles.
• 16:14
  SPEAKER_01: We're also talking physical infrastructure.
• 16:16
  SPEAKER_01: Picture AI agents watching pipelines,
• 16:19
  SPEAKER_01: spotting leaks or anomalies long before they turn
• 16:21
  SPEAKER_01: into major environmental disasters.
• 16:23
  SPEAKER_01: Or monitoring pump jacks,
• 16:24
  SPEAKER_01: fine-tuning their strokes per minute based on reservoir conditions and power costs.
• 16:28
  SPEAKER_00: Now you're speaking my language.
• 16:30
  SPEAKER_00: That's real-world impact.
• 16:31
  SPEAKER_00: Because a pipeline leak, that's a nightmare.
• 16:34
  SPEAKER_00: And a pump jack running inefficiently, that's leaving money right in the ground.
• 16:38
  SPEAKER_00: If an AI helps with that, then yeah, I'm listening.
• 16:40
  SPEAKER_00: That's not hype.
• 16:41
  SPEAKER_00: That's smart business.
• 16:42
  SPEAKER_01: And if an agent's going to touch a pipeline or pump jack,
• 16:45
  SPEAKER_01: the operator has to know why it made the call.
• 16:48
  SPEAKER_01: Eric Ho, co-founder and CEO of Goodfire, has been working on that problem.
• 16:52
  SPEAKER_01: Opening up the neural network instead of accepting the old black box answer.
• 16:56
  SPEAKER_00: The black box.
• 16:58
  SPEAKER_00: That's always been my concern.
• 16:59
  SPEAKER_00: You tell me the AI says do X, but you can't tell me why it says do X.
• 17:04
  SPEAKER_00: That makes me nervous, particularly for drilling decisions or safety protocols.
• 17:08
  SPEAKER_00: I need the reasoning.
• 17:09
  SPEAKER_01: Goodfire wants to make that possible.
• 17:11
  SPEAKER_01: They advocate for a more inspectable, debuggable approach to AI design.
• 17:16
  SPEAKER_01: Instead of only an output, you can trace back the AI's decision-making process.
• 17:20
  SPEAKER_01: You see which data points influenced which parts of the model to arrive at that conclusion.
• 17:24
  SPEAKER_00: You mean we can open the hood and see the engine running?
• 17:26
  SPEAKER_00: Not just hear it hum.
• 17:27
  SPEAKER_01: You've got the picture.
• 17:29
  SPEAKER_01: This proves critical for adoption in industries like ours,
• 17:32
  SPEAKER_01: where safety, compliance, and accountability are paramount.
• 17:35
  SPEAKER_01: If an AI recommends a certain well placement or a specific completion technique,
• 17:39
  SPEAKER_01: operators must understand why that recommendation is made.
• 17:42
  SPEAKER_01: They need to audit it, explain it, and stand behind it.
• 17:45
  SPEAKER_00: You can't tell the drilling engineer the magic box said so.
• 17:48
  SPEAKER_00: That won't fly.
• 17:49
  SPEAKER_00: Not with millions of dollars and potentially lives on the line.
• 17:52
  SPEAKER_00: You need to understand the logic.
• 17:53
  SPEAKER_00: You need to defend the decision.
• 17:55
  SPEAKER_00: If this interpretability thing actually works, it overcomes a major hurdle.
• 17:59
  SPEAKER_00: It removes some mystery, making it more trustworthy.
• 18:02
  SPEAKER_01: Building trust is key.
• 18:03
  SPEAKER_01: It also helps with debugging.
• 18:05
  SPEAKER_01: If an AI makes a bad recommendation or produces an unexpected result,
• 18:08
  SPEAKER_01: interpretability tools help engineers figure out why it happened so they can fix
• 18:12
  SPEAKER_01: the model.
• 18:13
  SPEAKER_01: This makes AI more reliable and understandable, not just smarter.
• 18:16
  SPEAKER_00: It's like a good landman.
• 18:18
  SPEAKER_00: You don't tell me the title's clean.
• 18:19
  SPEAKER_00: You show me the chain.
• 18:21
  SPEAKER_00: You show me the documents.
• 18:22
  SPEAKER_00: You show me the probate records, the affidavits.
• 18:24
  SPEAKER_00: You show me the work that builds trust.
• 18:27
  SPEAKER_01: That's the standard.
• 18:28
  SPEAKER_01: Show the work, then show the result.
• 18:30
  SPEAKER_01: Sounds right at home on Silicon & Shale.
• 18:32
  SPEAKER_00: Always a solid talk, Pack.
• 18:33
  SPEAKER_00: Good to chew on these topics.
• 18:34
  SPEAKER_00: Get them out in the open.
• 18:36
  SPEAKER_01: And that's our week.
• 18:37
  SPEAKER_01: Power, paper records, and machines that still have to show their work.
• 18:40
  SPEAKER_00: We'll be back next week.
• 18:41
  SPEAKER_00: Expect more insights.
• 18:42
  SPEAKER_00: Maybe another Courthouse Chronicle.
• 18:44
  SPEAKER_01: One last thing.
• 18:45
  SPEAKER_01: We're looking for sponsors who actually work in energy or technology and want
• 18:49
  SPEAKER_01: to reach the people doing the work.
• 18:50
  SPEAKER_01: If that sounds like your shop, email siliconandshale at gmail.com.
• 18:54
  SPEAKER_00: So for Packer, I'm Derek.
• 18:58
  SPEAKER_00: Keep your boots dry and your title clean.
• 19:00
  SPEAKER_01: Keep building everyone.