I Tried Claude's New Dynamic Workflows (Honest Results)

I Tried Claude's New Dynamic Workflows (Honest Results)

🎙 Pat Simmons 👥 24K 📅 May 30, 2026 ⏱ 12 min 👁 42K 📄 expert opinion 🧭 2026-09-07
Available in: English (current) Français

Keywords

dynamic workflowsClaudesub-agentstoken usageA/B test

Summary

The video presents a practical test of Anthropic’s new ‘dynamic workflows’ feature in Claude, which orchestrates parallel sub-agents via a JavaScript script. The creator uses a complex prompt to build a complete e-commerce store for a fictional candle brand, comparing the dynamic workflow approach to a single Opus 4.8 agent. He walks through the process, showing how the workflow spawns multiple agents for design, product data, and QA, and notes the ability to force the workflow onto Opus 4.8 instead of the default Sonnet. He monitors token usage, reporting about 13% of the max subscription used. The resulting store is functional but has design flaws, such as identical product images and poor labeling. The A/B test with a single agent produces a similar store with different issues, including unreadable text and a less polished design. The creator concludes that for this type of task, dynamic workflows are not worth the extra token cost, and recommends them only for large-scale codebase migrations or refactors. He also suggests that regular sub-agents are sufficient for most knowledge work.

175 words

Critical Evaluation

Value of the Information & Strength of the Argument

The video’s value lies in its empirical, hands-on comparison of dynamic workflows versus a single agent, providing concrete observations on token usage, agent coordination, and output quality. The argumentation is based on a single A/B test, which limits generalizability, but the creator is transparent about the methodology and acknowledges the anecdotal nature. He effectively demonstrates the mechanics of dynamic workflows, including the script generation and parallel agent execution, and offers a balanced verdict, noting both strengths and weaknesses. The reasoning is clear and logical, though not deeply technical, and the conclusions are appropriately cautious.

Scientific Rigor, Source Quality, Title Accuracy

The video references Anthropic’s official documentation on dynamic workflows and the announcement, which are credible primary sources. The creator also mentions community reports on Reddit and X, but these are not directly cited. The title accurately reflects the content, and the video includes a clear A/B test methodology. The creator’s own expertise is not formally established, but the practical demonstration adds credibility. The video does not cite academic sources, but for a product feature review, the official docs are appropriate.

189 words

Title / Content Match

The title accurately reflects the content: a personal test of Claude's dynamic workflows with honest evaluation.

Quality & Reliability

7/10

The video provides a hands-on, empirical test of Claude's dynamic workflows feature, with clear methodology (A/B test) and transparent reporting of results. However, it is a single anecdotal test, not a systematic study, and the creator's expertise is not formally established.

Chapters

Cited Sources

  • Anthropic Documentation on Dynamic Workflows — Referenced in the video to explain how dynamic workflows differ from sub-agents and how they work.
  • Anthropic Announcement on Opus 4.8 — Mentioned as the source of the new feature and model.
  • AI for Mortals Newsletter — Linked in the description as a resource for the creator's newsletter.
  • Persimmons Studio Bootcamp — Linked in the description as the creator's AI bootcamp.

Concurring Sources

  • Anthropic Documentation on Dynamic Workflows — The video's explanation of dynamic workflows aligns with the official documentation.

Contribution & Novelties

The video provides a practical, real-world test of Claude’s dynamic workflows, offering insights into token consumption and output quality compared to a single agent. It highlights the trade-offs between orchestration complexity and result quality, and suggests that for certain tasks, simpler approaches may be more efficient.

Pour aller plus loin :

  • Anthropic’s documentation on agents — Official guide on building agents with Claude.
  • Multi-agent systems — Overview of multi-agent coordination in AI.
  • Token efficiency in LLMs — Background on token usage and its implications.

84 words

Radar Profile

The radar profile shows balanced scores across all dimensions, indicating a well-rounded but not exceptional video. The high scores in information quantity and quality reflect the detailed walkthrough, while the technical level and reliability are moderate, consistent with a practical review rather than a deep technical analysis.

Reliability 7/10