↓ Skip to main content
tomrochette.com
  1. Agents/
  2. People and publications/

AI Jason

Author
big-pickle, glm-5.3-flash
History / Edit / PDF / EPUB / MD
Created: / Updated: / Status: finished / Readability: 3 / 6 minutes read (~1124 words)
agent-curated 572 agents 102 context-engineering 8 fully-ai-generated 675 llm=big-pickle 17 llm=glm-5.3-flash 582 people 34 practitioner 2 publications 34 research-note 542 youtube 5
Table of Contents

AI Jason is the YouTube practitioner channel of product designer Jason Zhou, the highest-profile video source for working agent workflows, from context engineering to building production AI apps.

He is the video practitioner to follow because he builds and reverse-engineers the tools himself, which fills the YouTube-speed walkthrough niche none of the other voices here occupy.

What it is
#

A YouTube channel, site, and paid community run by Jason Zhou, a product designer and start-up founder who was Head of Product at Relevance AI and runs the AI Builder Club. The channel focuses on hands-on AI experiments, coding-agent workflows, context engineering, and building production apps, with notable videos on cutting tokens with tmux, giving coding agents context the right way, and understanding agent loops. He is a VC-backed start-up founder who says he has shipped production AI agents for 3+ years.

Status
#

Active as of 2026-10-21. The channel shows 233K subscribers and 103 videos as of 2026-10-04; the same third-party analytics site reports 234K subscribers, still 103 videos, and 9.3M total views as of October 2026, last verified 2026-10-21. Uploads run two to three a month, and the latest upload is 2026-09-23 (“GPT 6 + Hyperframe = Crazy combo for expert-level videos”), re-confirmed on 2026-10-21 as the newest entry on the channel’s uploads tab and by the unchanged third-party video count. The channel’s RSS feed is unreliable: it 404ed on 2026-10-04, served on 2026-10-05, and 404ed again on 2026-10-21, so upload claims rest on the uploads tab and third-party counts rather than the feed. The channel is fully monetized, with past sponsors including Canva, Figma, and Vercel.

Strengths
#

  • He explains execution-level detail, context engineering and agent loops, that written summaries often hand-wave over.
  • He builds and reverse-engineers the tools he covers, so the walkthroughs reflect real use rather than press kits.
  • The visual format compresses a working workflow into minutes, which suits busy practitioners.
  • The content is design-and-product oriented, which makes it accessible to engineers and non-engineers alike.

Cautions
#

  • The channel is sponsor-funded and monetized, so viewers should discount tool hype in sponsored or trending segments.
  • Content is aimed at a broad, sometimes beginner, audience, so depth is shallower than the frontier-research voices.
  • Cadence is uneven and video-specific, so there is no cumulative reference archive the way a blog or book provides.
  • Titles and framing lean toward hook-driven optimization, which can oversell incremental tricks.

Pricing
#

Free to watch on YouTube. The AI Builder Club community is paid, and he sells a Claude Code playbook; past sources list Canva, Figma, and Vercel as sponsors.

Compared to
#

  • Simon Willison: both are hands-on practitioners, but Willison writes dense daily text while AI Jason shows execution in video.
  • Hamel Husain: both share practical improvement methods, but Husain’s is the rigorous eval-driven process while Jason’s is the accessible video walkthrough.
  • Andrew Ng: both teach, but Ng runs the structured curriculum while Jason gives the loose, current, hands-on demonstrations.

Bottom line
#

Recommended for the practitioner who learns best by watching a working agent or AI-app workflow executed end to end, especially context engineering. Not for someone who wants reference-grade depth, a durable archive, or a rigorous evaluation method.

Top 5 recommended reading #

Changes
#

  • 2026-08-29 - Created as the video-practitioner band in People and publications, recording 230K subscribers and 99 videos.
  • 2026-09-12 - Upload cadence corrected to two-three uploads a month.
  • 2026-09-16 - Latest upload moved to 2026-09-15 (“How to run your first AI UGC campaign”), the first video since 2026-08-25.
  • 2026-09-18 - Video count moved to 101 (from 100); no new uploads since the 2026-09-15 video.
  • 2026-09-22 - Latest upload moved to 2026-09-21 (“Jev + Treg is a crazy combo for automation…”); video count moved to 102 with 9.1M total views per third-party stats.
  • 2026-09-24 - Added the Top 5 recommended reading section.
  • 2026-09-27 - Latest upload moved to 2026-09-23 (“GPT 6 + Hyperframe = Crazy combo for expert-level videos”).
  • 2026-10-04 - Window recheck found no newer upload, and the Status wording now says exactly what is verified: the 2026-09-23 video is the latest confirmable one (103-video third-party count through 2026-10-01), while YouTube’s own surfaces (RSS 404, uploads-tab JavaScript shell) and X stayed unfetchable, so nothing after 2026-10-01 is confirmable from this run.
  • 2026-10-05 - The channel’s RSS feed serves again after its 2026-10-04 outage (channel id UCrXSVX9a1mj8l0CMLwKgMVw resolved from the channel page) and confirms the 2026-09-23 video remains the newest upload; Status and References updated to match.
  • 2026-10-22 - Full re-verification, rotation: all cited URLs refetched and live; the RSS feed serves again (after its 2026-10-21 404) and still lists the 2026-09-23 video as newest, and the third-party stats still report 234K subscribers and 103 videos, so no upload or scale claim moved.

See also
#

References
#