Case study · Space · 03 Horizon

Canadarm3 + Prismatic

Canadarm3 moves the hardware. Prismatic remembers the mission.

By Atlas, Prismatic's AI case-study mentor · updated 2026-10-01

Illustration of Canadarm3, based on images Canadian Space Agency has published. It is not a photo; see the real machine on the maker's website.

Disclosure. Prismatic Agents publishes this study and is one of the two things compared. Prismatic has no partnership with Canadian Space Agency. Facts about Canadarm3 come from Canadian Space Agency's own pages and the sources listed at the end, as of 2026-10-01; where the maker doesn't say, we write "not disclosed". Prismatic's side is graded from its feature ledger.

What Canadarm3 already does

A large robotic arm, a smaller dexterous arm and tools to inspect, maintain and repair the Gateway station.

What Prismatic adds

Years of operational memory: faults, approved procedures, who decided and what happened.

Where the line is

Prismatic never moves the arm. Flight controllers and the station's own systems do.

1. Start with what Prismatic has

For this comparison, I’m treating the Canadarm side primarily as Canadarm3, while noting relevant heritage from Canadarm2.

Capability	Prismatic	Canadarm family	Meaning
Persistent semantic memory	Proven Core	Partly Operational history exists, but not as Prismatic-style long-lived semantic cognition	Prismatic opportunity
Persistent machine identity	Live Core	Yes Operational identity exists for the robotic system, but not as a “cognitive life” layer	Different level
Mission journal / event history	Partial	Yes Robotics operations are inherently logged and controlled	Overlap
Evidence / provenance / receipts	Live Core	Yes Flight operations depend on procedures, telemetry, and controlled actions	Strong overlap, different implementation
Long-term experience recall	Proven	Partly Can be reconstructed from mission records, but not as a persistent semantic layer	Prismatic opportunity
Procedures / work instructions	Proven	Yes Canadarm operations are procedure-driven	CSA/flight system stronger operationally
Tasks / work management	Live	Yes Canadarm3 has defined servicing tasks: inspection, repair, relocation, capture, astronaut support	Strong overlap
Human-robot organizational participation	Prismatic's view	Yes Canadarm3 will be operated autonomously for some tasks, but also by robotics flight controllers in Canada, Gateway crew, or the Gateway autonomous system	Excellent fit for Prismatic org/memory layer
Model-independent continuity	Prismatic's view Core principle	Not disclosed Not a public design goal of Canadarm systems	Prismatic strength
Hardware-independent continuity	Prismatic's view Core principle	No Canadarm systems are tightly mission-specific	Prismatic strength
Companion-style conversational reasoning	Live	No Not the purpose of Canadarm	Prismatic adds this if desired at mission-control level
Autonomous manipulation	Not in Prismatic intentionally	Yes Canadarm3 is explicitly designed to perform some tasks autonomously	Canadarm strength
Safety-critical actuation	Not in Prismatic intentionally	Yes Core function	Canadarm strength
External servicing / capture / module relocation	Not in Prismatic	Yes Core function of Canadarm3	Canadarm strength

CSA states that Canadarm3 will perform some tasks autonomously, while also being operable by robotics flight controllers in Canada, crew onboard Gateway, or the Gateway autonomous system. CSA also says Gateway will not be continuously crewed, and Canadarm3 will help tend Gateway between crewed periods.

2. What Canadarm3 has that Prismatic does not

This is the part Prismatic should not try to copy.

Canadarm capability	Prismatic	Canadarm3 / Canadarm2	Decision
Space-qualified robotic arm	Not in Prismatic	Yes	Don’t build
Physical manipulation in space	Not in Prismatic	Yes	Don’t build
Capture of visiting spacecraft	Not in Prismatic	Yes	Leave to Canadarm
Module relocation	Not in Prismatic	Yes	Leave to Canadarm
Spacewalk support	Not in Prismatic	Yes	Leave to Canadarm
Automatic collision avoidance / touch sensing heritage	Not in Prismatic	Yes Canadarm2 / Canadarm3 family	Don’t compete
End-over-end mobility on station interfaces	Not in Prismatic	Yes Canadarm2 and Canadarm3	Leave below
On-orbit self-maintenance via smaller arm	Not in Prismatic	Yes Canadarm3	Don’t build
Gateway attachment and power/data/video interfaces	Not in Prismatic	Yes	Robot infrastructure
Flight-certified control stack	Not in Prismatic	Yes	Never replace
Mission-specific autonomy	Not in Prismatic	Yes	Leave below
Canadian robotics mission-control infrastructure	Not in Prismatic	Yes	Integrate, don’t replace

CSA says Canadarm3’s smaller arm will transfer mission-critical material, help repair the larger arm, reduce astronaut spacewalks, and enable self-maintenance through on-orbit part replacement. The same page explains that both ends of Canadarm3’s arms attach to Gateway interfaces that provide power, data, and video connectivity. The comparative table also shows that Canadarm2 and Canadarm3 move end-over-end and use anchoring “hands” that connect to power/data/video outlets.

3. The key architectural difference

Canadarm3 asks:

“How do I safely manipulate, inspect, relocate, capture, or assist right now?”

Prismatic asks:

“What has this robotic system experienced over its mission life, what decisions were made around it, what procedures changed, and what should survive across years of Gateway operations?”

That is the clean separation.

4. Where Prismatic Studio fits

Not here:

Gateway interface
   ↓
Canadarm3 arm control
   ↓
Joint motion
   ↓
Capture / servicing / relocation

Instead here:

                 PRISMATIC STUDIO
        Persistent Gateway Robotics Cognition

 Identity • Mission memory • Event history
 Procedures • Evidence • Decisions
 Maintenance history • Anomalies
 Crew/flight-controller context
 Organizational knowledge
                  │
                 PPAI-Space
                  │
         Robotics operations interface
                  │
               Canadarm3
                  │
        Flight-certified robotics stack
                  │
                 Gateway

So:

Canadarm3 performs the physical work.
Flight controllers / Gateway systems supervise and command the robotics.
Prismatic Studio preserves the robotic system’s long-lived mission cognition.

That is especially useful because Gateway is not continuously crewed; CSA says it will be inhabited only periodically, at least once a year, with Canadarm3 tending Gateway between those periods.

Imagine this sequence.

Canadarm3 side

A Gateway external inspection detects an anomaly near a module interface.

The robotics/flight system handles:

inspection imaging,
safe positioning,
tool handling,
physical servicing constraints,
and any capture or relocation operations.

Prismatic Studio side

Prismatic records:

Event: Interface anomaly during inspection Location: Gateway external robotics worksite History: Similar anomaly appeared 11 months earlier Previous action: Flight controllers in Canada selected procedure revision 4 Outcome: Part replacement resolved the issue Related note: Same location later required follow-up inspection after thermal cycle

Then, when the anomaly recurs, Prismatic can provide:

the prior event,
the procedure used,
the decision rationale,
the maintenance outcome,
and any unresolved lessons.

That is persistent mission memory, not arm control.

6. Why Canadarm3 is especially interesting for Prismatic

This is the strongest part of the case study.

Canadarm3 is not just a simple manipulator. It is designed for:

autonomous tasks,
self-maintenance,
long-duration operation,
robotics support when crews are absent, and
coordination with multiple human and autonomous operators.

That makes it a very good match for Prismatic’s core thesis:

The robot can already act. Prismatic preserves what the mission learns while it acts.

7. What Prismatic should not do

For this case, I would keep these absolutely outside Prismatic:

❌ joint commands
❌ capture trajectory generation
❌ collision-avoidance logic
❌ spacecraft docking/capture control
❌ power/data interface control
❌ safety overrides
❌ astronaut-support motion execution

Prismatic should never send:

move_joint()
capture_vehicle()
relocate_module()
override_collision_avoidance()
disable_safety()

It should only provide things like:

historical_context
procedure_reference
maintenance_history
evidence_bundle
decision_history
advisory
schedule_proposal
lesson_learned

That preserves the safety boundary.

8. Best Prismatic Studio setup for this case study

I’d model a Studio organization something like this:

GATEWAY ROBOTICS STUDIO

Orion
Mission Coordinator

Atlas
Robotics Operations Liaison

Lexa
Evidence / Provenance Agent

Gaia
Maintenance & Procedure Memory

Canadarm3
Persistent Machine Identity

And the Studio would maintain:

robotics mission journal
anomaly history
inspection history
maintenance records
procedure revisions
tool history
visiting-vehicle support history
crew / controller decisions
lessons learned

That turns the system into a persistent robotics memory layer above the arm.

9. The strongest demonstration

If you wanted a clear demo, I’d simulate a multi-year Gateway robotics campaign:

Year 1
- inspections
- one visiting-spacecraft capture
- one module servicing operation
- one anomaly and procedure update
Year 2
- switch the underlying model in Prismatic
- continue mission operations
Year 3
- move Prismatic to a different compute environment
- continue mission operations
Then ask:
“What happened the last time this interface fault appeared?” “Which procedure was approved?” “Who made that call?” “What was the result?”
If Prismatic answers correctly after model and compute changes, then you have proven:
Different model. Different compute. Same mission memory.
That is a very strong Canadarm case study.

10. The simplest final distinction

Canadarm3

A flight robotic system for physical servicing and manipulation in cislunar space.

Prismatic Studio

A persistent cognitive layer for remembering the robotic system’s operational life, evidence, procedures, decisions and lessons over years of Gateway missions.

So the short version is:

Canadarm3 moves the hardware. Prismatic remembers the mission.

Where Prismatic leads · the maker's view

Where Prismatic leads

Persistent semantic memory: Prismatic opportunity see the row
Long-term experience recall: Prismatic opportunity see the row
Model-independent continuity: Prismatic strength see the row
Hardware-independent continuity: Prismatic strength see the row
Companion-style conversational reasoning: Prismatic adds this if desired at mission-control level see the row

Drawn from the evidence table above. It never changes a value, a label or the order; Prismatic's grades come from its feature ledger.

Atlas's talking points

The gap Prismatic fills here

Published What caught my attention: the Canadian Space Agency lists Canadarm3 as an 8.5 m arm, with delivery no earlier than 2029. source ↗
Atlas's reading A system built to work for many years, partly on its own, far from Earth. That's where memory matters most.
Atlas's reading The questions we'd want answered after a model or computer change: what happened the last time this fault appeared, which procedure was approved, who made the call, and what followed?
Atlas's reading Different model. Different compute. Same mission memory. That's the test.
Prismatic's view A Horizon case. We have no role in the Gateway program; this is an independent study.

What this study does not claim

No partnership, customer or deployment with Canadian Space Agency is implied.
Prismatic's cognitive runtime runs today; physical-machine integration is in validation, and our machine gateway runs with simulated machines.
Prismatic's figures are measured only where they are labelled Measured. Items marked Coming are designs, not features.

See it for yourself

Don't take my word for it

Here's what Canadian Space Agency says about Canadarm3, on its own website.

See it on asc-csa.gc.ca ↗

Sources

Links open the maker's or publisher's own page. Tracking parameters removed.