E039: Starship Flight 5 and Tesla We, Robot

What E039 Saw in SpaceX and Tesla's October 2024 Demonstrations

E039 opened with a blunt question: what did you get done this week?

I used it to connect two October 2024 events. SpaceX flew Starship Flight 5 and caught the returning Super Heavy booster at the launch tower. Tesla presented Cybercab, Robovan, and Optimus at We, Robot. Both were visually ambitious. They did not provide the same kind of evidence.

flowchart LR
    A["October 2024 public event"] --> B["SpaceX Flight 5"]
    A --> C["Tesla We, Robot"]
    B --> D["Instrumented flight sequence"]
    D --> E["Achieved test objectives and remaining gates"]
    C --> F["Controlled rides, concepts, interactions, and forecasts"]
    F --> G["Product states and missing evidence"]

Flight 5 made a specific sequence observable

SpaceX's fifth Starship flight test launched on October 13, 2024. The company's mission record reports ascent, hot-stage separation, Super Heavy boostback, a landing burn, and the first catch of the booster by the launch and catch tower.

The upper stage continued across its planned trajectory, reentered through peak heating and aerodynamic pressure, performed a flip and landing burn, and splashed down in the target area of the Indian Ocean.

The catch clip dominated the conversation because it compressed a difficult control and ground-system problem into a few visible seconds. I became fixated on the catch hardware and the coordination between booster and tower.

The rigorous version of that excitement is still strong. A flight article completed an integrated return and catch sequence under flight conditions. SpaceX says thousands of vehicle and pad criteria had to pass before the catch attempt.

That does not establish every conclusion I connected to it. One successful catch does not prove rapid refurbishment, repeated reliability, crew safety, daily launch cadence, manufacturing rate, payload economics, orbital refueling, lunar operations, or Mars transport.

The FAA's Flight 5 activity record explains the environmental re-evaluation used for the license modification. Authorization defined a regulatory boundary. It did not certify every engineering claim.

The catch was an achievement and a beginning

The tower catch moves part of the recovery problem from landing hardware on the booster to a coordinated vehicle and ground system. The booster still has to complete separation, boostback, coast, landing burn, and approach before the arms receive it.

Public sources do not reveal the full control law, tolerance, structural margin, or commit criteria. That is why the explainer should stop at a sourceable system level.

The next evidence comes from inspection, repeated attempts, different conditions, anomalies, design changes, maintenance, and actual reuse. A milestone can be historically important without being a reliability distribution.

Tesla's event showed three different product states

Tesla's We, Robot event page presented Cybercab, Robovan, and Optimus together. The staging encouraged one story about autonomous transport and robotics.

The evidence needs three records.

Cybercab was a purpose-built vehicle concept without conventional driver controls. Attendees experienced controlled rides at the event. Robovan was presented as a larger passenger or cargo concept. Optimus units walked and interacted with attendees in a prepared venue.

The event also included company forecasts about production, timing, cost, and capability. A forecast is not a result merely because it appeared next to a working prototype.

I was especially interested in control mode. The Optimus interactions looked persuasive, but I had seen reports and attendee observations suggesting human involvement. I did not have enough verified evidence in the episode to state exactly which actions or voices were autonomous, scripted, teleoperated, remotely assisted, or safety-controlled.

That uncertainty should stay visible. A humanoid robot can walk autonomously while a remote person supplies speech. A person can set a high-level goal without controlling every joint. A safety operator can intervene without operating continuously. Appearance alone cannot tell us which mode was active.

Later company filings changed the state without erasing the event

Tesla's 2025 Form 10-K states that the company launched a Robotaxi service using Model Y vehicles in June 2025 and expects Cybercab to join later.

Tesla's January 2026 company update lists Cybercab as tooling and an Optimus facility as construction, with later production plans. The document also warns that installed capacity is not a current production rate.

Those updates matter. They do not mean Cybercab was in customer production at the 2024 event. They do not establish Robovan production. They do not prove autonomous general-purpose Optimus work.

The maintained record should name the product, date, evidence, company claim, regulator state, production state, service state, and unresolved question.

Automated driving needs a current driver-role answer

In E039, I described Tesla Full Self-Driving as not yet full self-driving. The current product name now makes the boundary more visible.

Tesla's current FSD Supervised support page says the system requires an attentive driver and does not make the vehicle autonomous. NHTSA's automation guide distinguishes Level 2 driver assistance, where the driver remains responsible, from an automated driving system that performs the full driving task within a defined domain.

The practical question is not whether a brand says “self-driving.” It is who drives, who monitors, who performs fallback, where the system operates, what permit applies, and what current evidence supports the claim.

A better definition of execution

Execution is not the number of dramatic objects on a stage. It is movement through declared evidence states.

Flight 5 moved a specific integrated flight-test objective through a successful observed sequence. The result still needed inspection, repetition, and operations.

We, Robot revealed three concepts and gave attendees controlled experiences. Each product then needed its own safety, permit, production, service, and autonomy record.

The question “what did you get done?” becomes more useful when the answer includes what was tested, what passed, what remained forecast, and what evidence would move the claim again.

This article was developed with AI assistance from the immutable E039 transcript and linked SpaceX, FAA, Tesla, NHTSA, product-state, and historical-boundary records. Dalton Anderson remains the author. Transcript, aerospace, vehicle-safety, robotics, current-source, and founder review are mandatory before publication. Publication is not authorized.