Multi-Turn Reasoning: Agents That Think Across Steps

Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection, critique, retry — that give you AI which actually solves hard problems.

40 min · Reviewed 2026

Beyond one-shot prompting

A single prompt forces the AI to generate an answer in one go. For hard problems — proofs, long code, research — even the best model gets some parts wrong on the first pass. Multi-turn flows let the model reflect, critique itself, and retry.

The generate-critique-revise loop

TURN 1 (generate):
'Write a 200-word argument for why schools should start later. Make it persuasive.'
--> draft v1

TURN 2 (critique):
'You are a skeptical peer reviewer. List the three weakest points in the argument above. Be specific.'
--> critique list

TURN 3 (revise):
'Using the critique, revise the argument. Address each of the three weaknesses. Keep it under 200 words.'
--> draft v2A classic three-turn improvement cycle.

Each turn plays a different role — author, critic, editor. Studies on Claude and GPT consistently show that a two- or three-pass structure outperforms a single pass, especially on reasoning-heavy tasks. The model is effectively its own reviewer.

The ReAct pattern (reason + act)

ReAct interleaves reasoning steps with tool calls. Instead of one big plan, the model thinks a little, uses a tool (search, calculator, code execution), reads the result, thinks again, and so on. This is the foundation of modern AI agents.

THOUGHT: I need to find the CEO of Notion and their hiring trends.
ACTION: search("Notion CEO")
OBSERVATION: Ivan Zhao is CEO of Notion Labs.

THOUGHT: Now I need recent hiring data for Notion.
ACTION: search("Notion hiring 2026")
OBSERVATION: Notion announced 200 new engineering roles in Q1 2026.

THOUGHT: I have enough to answer.
FINAL: Notion, led by CEO Ivan Zhao, announced 200 new engineering roles in Q1 2026.ReAct loop. Thought / Action / Observation repeats until Final.

Reflection on tool failures

If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
If a tool returns empty results, try a broader query before giving up.
If results contradict each other, surface the contradiction rather than silently picking one.
Cap the loop — set a max of N iterations so the agent can't spin forever.

Keeping context manageable

Multi-turn flows accumulate tokens quickly. Strategies: summarize older turns into a running memo; store long data in a scratchpad tool; use XML tags to section the conversation. A 1M-context Claude can hold a lot, but cost and latency still grow.

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-prompting-multi-turn-creators

What is the core idea behind "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection, critique, retry — that give you AI which actually solves hard problems.
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which term best describes a foundational idea in "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. ReAct
2. multi-turn reasoning
3. agent loop
4. reflection
A learner studying Multi-Turn Reasoning: Agents That Think Across Steps would need to understand which concept?
1. multi-turn reasoning
2. agent loop
3. ReAct
4. reflection
Which of these is directly relevant to Multi-Turn Reasoning: Agents That Think Across Steps?
1. multi-turn reasoning
2. ReAct
3. reflection
4. agent loop
Which of the following is a key point about Multi-Turn Reasoning: Agents That Think Across Steps?
1. If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
2. If a tool returns empty results, try a broader query before giving up.
3. If results contradict each other, surface the contradiction rather than silently picking one.
4. Cap the loop — set a max of N iterations so the agent can't spin forever.
Which of these does NOT belong in a discussion of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. If results contradict each other, surface the contradiction rather than silently picking one.
3. If a tool returns empty results, try a broader query before giving up.
4. If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
What is the recommended tip about "Practitioner tip" in the context of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Treat every prompt as a spec: role → context → task → format. Review your first output as a draft, not a final.
4. Suggest CoT for math, planning, multi-step logic.
Which statement accurately describes an aspect of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Suggest CoT for math, planning, multi-step logic.
4. A single prompt forces the AI to generate an answer in one go. For hard problems — proofs, long code, research — even the best model gets so…
What does working with Multi-Turn Reasoning: Agents That Think Across Steps typically involve?
1. Each turn plays a different role — author, critic, editor. Studies on Claude and GPT consistently show that a two- or three-pass structure o…
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which of the following is true about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. ReAct interleaves reasoning steps with tool calls. Instead of one big plan, the model thinks a little, uses a tool (search, calculator, code…
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which best describes the scope of "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. It is unrelated to prompting workflows
2. It applies only to the opposite beginner tier
3. It focuses on Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Suggest CoT for math, planning, multi-step logic.
4. The generate-critique-revise loop
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. The ReAct pattern (reason + act)
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Reflection on tool failures
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Keeping context manageable
4. Suggest CoT for math, planning, multi-step logic.

← Back to interactive lesson

Tendril · Creators · Prompting

Multi-Turn Reasoning: Agents That Think Across Steps

Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection, critique, retry — that give you AI which actually solves hard problems.

40 min · Reviewed 2026

Beyond one-shot prompting

The generate-critique-revise loop

TURN 1 (generate):
'Write a 200-word argument for why schools should start later. Make it persuasive.'
--> draft v1

TURN 2 (critique):
'You are a skeptical peer reviewer. List the three weakest points in the argument above. Be specific.'
--> critique list

TURN 3 (revise):
'Using the critique, revise the argument. Address each of the three weaknesses. Keep it under 200 words.'
--> draft v2A classic three-turn improvement cycle.

The ReAct pattern (reason + act)

THOUGHT: I need to find the CEO of Notion and their hiring trends.
ACTION: search("Notion CEO")
OBSERVATION: Ivan Zhao is CEO of Notion Labs.

THOUGHT: Now I need recent hiring data for Notion.
ACTION: search("Notion hiring 2026")
OBSERVATION: Notion announced 200 new engineering roles in Q1 2026.

THOUGHT: I have enough to answer.
FINAL: Notion, led by CEO Ivan Zhao, announced 200 new engineering roles in Q1 2026.ReAct loop. Thought / Action / Observation repeats until Final.

Reflection on tool failures

If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
If a tool returns empty results, try a broader query before giving up.
If results contradict each other, surface the contradiction rather than silently picking one.
Cap the loop — set a max of N iterations so the agent can't spin forever.

Keeping context manageable

End-of-lesson check

15 questions · take it digitally for instant feedback at tendril.neural-forge.io/learn/quiz/end-prompting-multi-turn-creators

What is the core idea behind "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection, critique, retry — that give you AI which actually solves hard problems.
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which term best describes a foundational idea in "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. ReAct
2. multi-turn reasoning
3. agent loop
4. reflection
A learner studying Multi-Turn Reasoning: Agents That Think Across Steps would need to understand which concept?
1. multi-turn reasoning
2. agent loop
3. ReAct
4. reflection
Which of these is directly relevant to Multi-Turn Reasoning: Agents That Think Across Steps?
1. multi-turn reasoning
2. ReAct
3. reflection
4. agent loop
Which of the following is a key point about Multi-Turn Reasoning: Agents That Think Across Steps?
1. If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
2. If a tool returns empty results, try a broader query before giving up.
3. If results contradict each other, surface the contradiction rather than silently picking one.
4. Cap the loop — set a max of N iterations so the agent can't spin forever.
Which of these does NOT belong in a discussion of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. If results contradict each other, surface the contradiction rather than silently picking one.
3. If a tool returns empty results, try a broader query before giving up.
4. If a tool errors, don't just retry blindly — have the model reason about why and adjust inputs.
What is the recommended tip about "Practitioner tip" in the context of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Treat every prompt as a spec: role → context → task → format. Review your first output as a draft, not a final.
4. Suggest CoT for math, planning, multi-step logic.
Which statement accurately describes an aspect of Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Suggest CoT for math, planning, multi-step logic.
4. A single prompt forces the AI to generate an answer in one go. For hard problems — proofs, long code, research — even the best model gets so…
What does working with Multi-Turn Reasoning: Agents That Think Across Steps typically involve?
1. Each turn plays a different role — author, critic, editor. Studies on Claude and GPT consistently show that a two- or three-pass structure o…
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which of the following is true about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. ReAct interleaves reasoning steps with tool calls. Instead of one big plan, the model thinks a little, uses a tool (search, calculator, code…
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which best describes the scope of "Multi-Turn Reasoning: Agents That Think Across Steps"?
1. It is unrelated to prompting workflows
2. It applies only to the opposite beginner tier
3. It focuses on Some problems need more than one prompt. Learn how to design multi-turn reasoning flows — reflection
4. It was deprecated in 2024 and no longer relevant
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Suggest CoT for math, planning, multi-step logic.
4. The generate-critique-revise loop
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. The ReAct pattern (reason + act)
2. Track units through chained operations reliably.
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Reflection on tool failures
3. Build thumbs-up/down or rating mechanisms in user-facing AI
4. Suggest CoT for math, planning, multi-step logic.
Which section heading best belongs in a lesson about Multi-Turn Reasoning: Agents That Think Across Steps?
1. Track units through chained operations reliably.
2. Build thumbs-up/down or rating mechanisms in user-facing AI
3. Keeping context manageable
4. Suggest CoT for math, planning, multi-step logic.

← Back to interactive lesson