Once an agent starts generating, the only intervention is a hard stop — and stopping throws away whatever was already produced.
← Back to patterns
Agent Workflow
Mid-Stream Steering
Talk to the AI while it works
Don't stop the agent — steer it. While the AI is streaming, the user can drop quick directives ("shorter", "less formal", "go deeper on the second point"). The output keeps flowing, but with the new direction baked in, marked inline so the pivot is visible. No restart, no wasted tokens.
Metaphor
Steering Wheel
Interaction
Live Redirect, Quick Nudge
Use case
Long-form Generation, Agent Reasoning, Streaming Outputs
Failure mode
Atomic Generation
Control mechanisms
Steering Chips · Inline Course Marker · Soft Pivot (no restart)
Reference examples
ChatGPT Voice · OpenAI Deep Research mid-run Update · Claude Code mid-turn messages · OpenAI Codex Steer
Concept Frame
Stop is not the same as steer.
Stop buttons are an emergency brake — they throw away in-progress work and force the user to start over. That is a binary decision in a domain that wants nuance: most of the time you don't want the AI to stop, you want it to turn slightly. Mid-Stream Steering treats generation as a moving conversation, with the output as a stream you can lean on rather than a finished artefact you can only accept or reject.
Interactive Example
The live side-by-side is the comparison.
This prototype already contains the contrast between chat and the pattern response. It stands as the primary comparison surface for this page.
Today the only brake is a hard stop.
Once an agent starts generating, chat offers exactly one intervention — stop — and stopping throws away everything already produced. So the user waits out a stream they can already tell is going wrong, reads it to the end, then asks for a redo from scratch. Generation is treated as atomic, which turns every course-correction into wasted tokens.
Mid-Stream Steering lets the user drop quick directives while the output flows — "shorter," "less formal," "go deeper on the second point" — and the stream continues with the new direction baked in and marked inline at the pivot. No restart, no discarded work, and every steered segment is annotated with the directive that caused it. Redirection moves from after the work to during it.
"The cheapest course-correction is the one you make while the car is still moving, not after you've parked and started over."
DOPE Lens
How this pattern improves AI interaction quality.
Observability
Course-correction markers appear inline so the pivot is visible in the output itself
Directability
Direct manipulation of in-flight generation — change tone, depth, or angle without stopping
Predictability
Each steering chip describes the new direction in plain language before applying
Explainability
Every steered segment is annotated with the directive that caused it
Fit
Where it earns its place.
Use it
when
when
Generations are long enough that waiting and redoing costs more than redirecting in flight.
Avoid
when
when
Outputs are short enough that waiting and redoing is cheaper than building a steering surface.
In the wild
Where this pattern already ships.
Real implementations from products users already know. Each entry shows how the pattern got applied — what shipped, where it sits in the product, what it traded off.
ChatGPT Voice
OpenAI
Voice's Live option listens while it speaks, so you can cut in mid-answer and it should follow the latest thing you said — the cleanest consumer form of soft mid-stream steering. The help page names the limits: overlapping speech and background noise can blur what it hears.
Visit OpenAI ↗
OpenAI Deep Research mid-run Update
OpenAI
Feb 2026: users add new info mid-research via sidebar "Update" button. No restart, progress preserved. The non-destructive pivot the pattern actually describes.
Visit OpenAI ↗
Claude Code mid-turn messages
Anthropic
Sep 2026 docs: a message typed while Claude runs tool calls reaches it as soon as those calls finish, inside the same turn — the soft pivot, no restart. The steered turn shows the message where it landed; there is no per-segment course marker.
Visit Anthropic ↗
OpenAI Codex Steer
OpenAI
Captured Sep 2026: Steer pushed “Skip mercury baths, and name the inventor for each step” into a running five-search task. It lands in the timeline between two searches, the agent acknowledges it and keeps going — no restart. In a pure text stream with no step boundary the same message only lands when the turn ends. The Responses API ships the primitive for builders as response.steer.
Visit OpenAI ↗
Research Frame
Why this pattern exists, and where it breaks.
Let users redirect the in-flight stream with quick chips or natural language; the output continues with the new direction marked inline.
Chat treats generation as atomic — you wait, read, and then ask for a redo. Mid-stream steering lets the redirection happen during the work, not after it.
Too many steering options can fragment output and produce zigzag voice; markers can clutter the result.
Common questions
What people ask about this pattern.
When should I use the Mid-Stream Steering pattern?
+
Generations are long enough that waiting and redoing costs more than redirecting in flight.
When should I avoid the Mid-Stream Steering pattern?
+
Outputs are short enough that waiting and redoing is cheaper than building a steering surface.
What problem does Mid-Stream Steering solve?
+
Once an agent starts generating, the only intervention is a hard stop — and stopping throws away whatever was already produced.
Why is chat the wrong fit for this?
+
Chat treats generation as atomic — you wait, read, and then ask for a redo. Mid-stream steering lets the redirection happen during the work, not after it.
Related Patterns
Curated connections, not just neighbors.
Hand-picked relationships — what this pattern pairs with in practice, what it replaces, what it builds on, and what builds on it.
Often paired with
Editable Plan
Editable Plan sets the course before kickoff; Mid-Stream Steering corrects it while the output flows.
Often paired with
Follow-up Queue
Mid-Stream Steering acts on the running turn; Follow-up Queue lines up asks for after it ends, and any card can steer it now.
Alternative to
Approval Gates
Approval Gates stop the run at set checkpoints for sign-off; Mid-Stream Steering redirects it without stopping at all.