Build1 distinct publisher3 min readUpdated
Mythos AI's towboat advisor and Aurora's driverless trucks are not rival technologies so much as rival evidentiary standards. Only one of them has a published number, and it is not a savings figure.
The Engineer · Build desk

Compiled by The EngineerSomething wrong?How this is made
Southern Devall installed Mythos AI's Advanced Pilot Assist System on a commercial Mississippi River towboat in 2025 while leaving control of the vessel with the crew [5][9], and Aurora Innovation is running Class 8 trucks in Texas with nobody behind the wheel [2]. The technical gap between those two matters less than the evidence gap: according to the reporting, Mythos has to show measured gains from a crew-assistance installation, while Aurora has to show revenue and reliable utilization from fully driverless operations [3]. Both approaches appear in an August 14 Reuters analysis of automation across US freight [4].
The assist bet is the harder one to prove, because the asset already works without it. APAS monitors hazards, tracks vessels, calculates stopping distances, analyzes river conditions and identifies fuel-saving strategies, Reuters reported [8]. The Waterways Journal, attributing its account to a Southern Devall statement, described a system that gives the captain information and alarms while the crew keeps control [6][9]. Southern Devall is a credible test bed: the company says it operates more than 70 towboats and 230 barges with about 700 mariners, from headquarters in Sulphur, Louisiana, and across Louisiana, Texas and Tennessee [10]. That works out to no more than about ten mariners per towboat, so the labour line is not what the software touches [26].
What is missing is the delta. Neither Mythos AI nor Southern Devall has published independently measured fuel savings or safety improvements from the installation [11]. The theoretical case runs through the underlying economics of the mode: a US Army Corps of Engineers comparison cited by Reuters found one gallon of fuel can move one ton of cargo more than 500 miles by barge against about 60 miles by truck [12], roughly an eightfold advantage per gallon [13]. Better stopping calculations and traffic planning could trim waste on top of that, but the share attributable to Mythos AI remains unquantified [14].
Mythos AI's own disclosed numbers are from its other product line. In September 2025 the company said its fully autonomous MNAV system had logged more than 11,000 miles and over 15,000 encounters compliant with international collision-avoidance rules, figures supplied by the company and not independently validated [15]. In November 2025 Ocean Power Technologies announced plans to integrate Mythos AI software with its WAM-V vehicles and PowerBuoy platforms, with first demonstrations scheduled for the first quarter of 2026; no authoritative source confirms those demonstrations happened [16]. In July 2026 BlackSea Technologies named Mythos AI the first autonomy partner for its Mainstay architecture [17]. Total funding, valuation and pricing are not public, and PortXL is the only backer identified [18]. The company was founded in 2020 by Geoff Douglass, John Houston and Allen Flick in West Palm Beach [22], which puts five years between incorporation and the first named commercial installation [23].
Aurora's clock is longer. Urmson co-founded the Pittsburgh company in 2017 after leading Carnegie Mellon's DARPA challenge teams and serving as chief technology officer of Google's self-driving program, with former Tesla Autopilot director Sterling Anderson among the co-founders [20]. Commercial freight began in April 2025 and driverless operations expanded across Sun Belt routes in 2026 [7]: about eight years from founding to first revenue [21].
Watch for a before-and-after fuel burn or incident number from the Southern Devall fleet, since the disclosure gap is what currently limits comparison with better-capitalized road-autonomy firms [19]. Watch whether the Ocean Power demonstrations were actually run [16]. And on the road side, watch for utilization, not mileage.
Follow any of these and your For You feed starts watching them — no settings page required.
Ranked by verification strength, evidence, and original report placement.
Geoff Douglass's Mythos AI entered commercial freight with software that advises a Mississippi River towboat crew while leaving the vessel under human control.
Chris Urmson's Aurora Innovation is deploying Class 8 trucks in Texas without anyone behind the wheel.
The distinction shapes what each founder must prove: Mythos needs measured gains from a crew-assistance installation, while Aurora needs revenue and reliable utilization from fully driverless operations.
The two approaches feature in an August 14 Reuters analysis of automation across US freight.
Reuters reported that Southern Devall installed Mythos AI's Advanced Pilot Assist System, or APAS, on a commercial Mississippi River towboat in 2025.
The Waterways Journal attributed its account of the installation to a Southern Devall statement.
Evidence-backed comparisons of source perspectives and observed adoption signals. Read the methodology
Which Builder, Operator, and Investor concerns the observed source mix emphasized—not a truth score.
Evidence, demonstrated adoption, hype gap, incentives, and confidence are assessed independently, each on its own current evidence. How these are measured.
Mixed: audited on one side, self-reported on the other
Aurora's operating and financial figures come from a quarterly shareholder letter with explicit period boundaries, which is comparatively strong evidence. The Mythos AI side rests on a wire and trade account whose underlying attribution is a Southern Devall statement, plus company-supplied autonomy metrics the covering source explicitly labels unvalidated, an announced integration whose demonstrations are unconfirmed, and no published performance measurement at all. Only one publisher covers the cluster, so there is no independent corroboration of the maritime installation.
Real revenue-bearing deployments, still narrow
Both models are past demonstration: one commercial towboat carries APAS inside a 70-plus-boat fleet, and Aurora has run nearly 440,000 driverless miles with $2 million of quarterly revenue and a stated year-end target above 200 trucks. But maritime adoption is a single installed vessel with two announced partner integrations (one unconfirmed), and Aurora's revenue against a $266 million quarterly operating loss shows commercial volume that is still immaterial relative to cost. Adoption is genuine and measurable, not yet broad.
Mildly overstated, but the coverage discounts itself
The residual overstatement sits with the primary actors rather than the coverage: an advisory system is presented against a fuel-efficiency backdrop (500-plus barge ton-miles per gallon versus about 60 by truck) whose savings share attributable to the software is unquantified, and autonomy mileage is promoted without validation. The covering article partly closes that gap by flagging the missing measurements, the unvalidated figures and the unconfirmed demonstrations, which is why the gap is small rather than large.
Promotional sourcing on the private side
Most maritime facts originate with parties that benefit from them: Mythos AI supplied the autonomy metrics, Southern Devall supplied the installation account, Ocean Power Technologies and BlackSea announced partnerships involving their own products, and the founder-and-founding-date detail comes from a venture firm's promotional American Dynamism 50 listing. Aurora's numbers are investor-facing disclosures, which carry their own presentational incentive around a forward fleet target and back-loaded revenue guidance, though with more accountability. Absent funding, valuation and pricing disclosure from Mythos AI, incentive-laden sourcing is doing heavy lifting.
Moderate: dates and disclosures firm, outcomes unmeasured
Structural facts are well specified and internally consistent, including named companies, dated announcements, fleet counts and quarterly financials with stated cutoffs. Confidence is held down by single-publisher coverage, month-only precision on several events, one confirmed maritime installation, and the complete absence of outcome measurement on the crew-assist side. The assessment of what each model has proved is more reliable than any judgment about whether either works.
build
California's heavy-truck permits start a mileage clock, not a driverless freight service1 distinct publisher
product
California's first heavy AV truck permits arrive with a lawsuit already attached1 distinct publisher
product
EliseAI's mark moves from $2.2bn to $3.7bn on tenant texts and tour bookings1 distinct publisher
invest
The regulated are now the biggest donors: $517m of corporate money into the 2026 midterms1 distinct publisher
Distinct publishers with included, body-backed reporting in this cluster.
1 article · August 14, 2026