Rez

curious guy

[Rez @allalexreznik]

@gabriel1 same with news. I'm still the reader and the paying sub. must I scroll myself?

[Rez @allalexreznik]

https://t.co/mQy2Q3XtvN

I should probably add sound effects

[Rez @allalexreznik]

@jeremygiffon yeah I'll follow this account

[Rez @allalexreznik]

@AlexHormozi @almhatre check number 2

[Rez @allalexreznik]

@almhatre this was identified early on as a probable outcome

[Rez @allalexreznik]

9/9
TL;DR: Google, please fix your damn anchored comments API so I can at least build this! This is my long winded support ticket. LMK

[Rez @allalexreznik]

8/9
And guess what? Google Docs anchored comments API doesn’t actually work like it says. It's documented but the comments don't show up. Total roadblock...

[Rez @allalexreznik]

7/9
It’s a sweet pipeline: @OpenAIDevs GPT-4 Vision sees highlights, @AssemblyAI transcribes my voice, and
@AnthropicAI Claude 3.5 writes the comments. Then I map everything back to the doc lines via timestamps.

[Rez @allalexreznik]

6/9
What's actually useful: recording a single read-through on Loom, using AI vision to track highlights, transcribing to capture speech with full document context to round them out, and an LLM to generate coherent anchored comments.

[Rez @allalexreznik]

5/9
So I tested a system with voice memos on highlighted sections—cool, but still clunky. i.e AI transcribed voice memo comments. You have to record, maybe redo if you slip up, and coordinate recordings. Too much button-clicking…

[Rez @allalexreznik]

4/9
Recording a Loom might seem like a fix, but it’s a headache for the writer. They have to slow your video, find timestamps, then copy-paste your comments. This just doesn't work for both sides

[Rez @allalexreznik]

2/9
@googledrive Docs API is painful but mostly usable—until you hit anchored comments, which everyone uses to tag specific text. Floating comments in the void? Pretty useless for any real editing above top-level feedback