Research Note
E033 Historical Product Experience Record
The YouTube English auto-caption file is the recording-derived source for E033. The Google Drive file is a production outline. The legacy article is a later editorial art
E033 Historical Product Experience Record
Recording boundary
The YouTube English auto-caption file is the recording-derived source for E033. The Google Drive file is a production outline. The legacy article is a later editorial artifact. Neither replaces the captions when describing what aired.
The recording was published in September 2024. It captures launch-era Gemini Gems and Meta AI Studio, a live comparison of two Curio configurations, Dalton's Venture Episode Engine, and a live travel-planner Gem demonstration.
Durable observations
Dalton found that persistent instructions were easier to reuse than repeatedly rebuilding a long chat. His podcast-outline example began with notes, voice input, and a preferred outline structure. A custom Gem turned that repeated method into a named configuration.
He also noticed that generated instructions still needed editing. The Episode Engine required several passes, its outline order drifted, and the output used more bold formatting than he wanted. The live travel planner omitted the fact that he would be working during the trip because he had not supplied that constraint.
Those moments support a bounded conclusion. Reusable instructions lower setup friction, but they do not discover missing requirements, prove expertise, or remove review.
Dated product observations
The recording describes Gemini Gems as a more open instruction canvas and Meta AI Studio as a more socially distributed character system. It also describes device access, discoverability, audience defaults, prompt lengths, and product behavior as Dalton observed them.
Those details are frozen observations, not current claims. Current Google documentation says Gems can be created in the web app, used in mobile and Workspace surfaces, supplied with knowledge files, and shared under several access levels. Current Meta help describes audience choices including Everyone, Close Friends, and Only Me. These later states correct the legacy article's current-tense claims without rewriting the recording.
Editorial correction
E033 uses "agent" broadly. The public package should preserve that historical language while explaining that most demonstrated systems were persistent instructions around a chat model. No demonstrated system independently initiated work, held broad permissions, executed a durable multi-step loop, or acted without a user prompt.
Personal and safety boundary
The black-hole exchange, travel plan, email-report example, and personal productivity estimates were demonstrations and opinions. They did not establish scientific accuracy, safe travel advice, authorized workplace data use, or measured productivity gains.
Sources
Follow the evidence.
- youtu.be: nAW62 6pXaUyoutu.be
- tsapps.nist.gov: get pdftsapps.nist.gov
- blog.google: google gemini update august 2024blog.google
- support.google.com: 15146780support.google.com
- about.fb.com: create your own custom ai with ai studioabout.fb.com
- NIST AI Risk Management Frameworknist.gov
- Gemini Apps Privacy Hubsupport.google.com
- privacycenter.instagram.com: policyprivacycenter.instagram.com
- daltonanderson.ghost.io: google gems vs meta ai building your first ai agentdaltonanderson.ghost.io
- ai.meta.com: ai studioai.meta.com
- facebook.com: 1675196359893731facebook.com
- support.google.com: 15235603support.google.com
- support.google.com: 16504957support.google.com
- open.spotify.com: 0ZMJAP0X2CzPVbC83gaWagopen.spotify.com