Yes they conveniently forget about disclosing prompt processing time. There is an affordable answer to this, will be open sourcing the design and sw soon.
Very much doubt, what you’ll see is it killing off paralegal work.
In most jurisdictions legal advice is a regulated and restricted activity. Qualified lawyers today get themselves into trouble without AI advising on areas they have no right to practice in.
For the average office task they don’t seem far off being competent, at least to the average workers quality.
Ai builder with gpt5 + workflow triggers is very capable already. 1-2 more model generation hops needed plus a bit more “agent” plumbing before its game over for the excel and word jobs.
Which average office tasks would those be? Writing project proposals? Putting budget numbers into a shared spreadsheet? Composing a progress report? Preparing presentation slides for an executive status update meeting? Writing performance reviews? Taking mandatory compliance training? Going to planning meetings?
One or two of these, I could see. Automated progress reports would be nice. But a lot of them aren’t about document generation, but about human accountability, about being a person who commits to something in writing. Automating away paper pushers means all the accountability lands on their bosses, leaving them nowhere to hide. It will be quite something if we manage to rewrite the corporate social context like this.
Gemini 3 would rip through that problem, but equally you could slice the video with existing open source tooling such as FFMPEG then combine with blender for the video curation. Gemini 3 could probably write you the workflow as well.
How many crashes per mile for Waymo etc. would be the more interesting metric - if the competition has better numbers there is no excuse for risking people's safety with inferior tech
Sorry don’t like the max model, feels like it needs a lot more guiding. The plans it writes however are better, so I tried feeding it back in (meta prompt style) and working okay so far. Very large repository.