AI can research, write, brainstorm, and pitch in seconds. But can it actually do your job—or just produce work that looks finished?
I put ChatGPT’s new Terra 6.1 model to the test by handing it five tasks from a typical knowledge worker’s day. The results were impressive, occasionally frustrating, and a useful reminder that doing work quickly isn’t the same as doing it well.
Five Tasks. One AI. Almost No Waiting.
OpenAI’s latest model, Terra 6.1, promises faster computer-based work, stronger video analysis, and better efficiency. Naturally, I wanted to know what that meant in practice.
As a journalist, founder, and podcast host, my working day involves plenty of moving parts: researching news, writing articles, preparing interviews, growing an audience, and developing commercial partnerships.
So I gave Terra 6.1 five assignments:
Find three breaking news opportunities and verify them.
Research and draft an 800-word article using primary sources.
Write ten original interview questions for a technology executive.
Create a LinkedIn post, newsletter teaser, and distribution plan.
Identify five potential commercial partners and draft personalized pitches.
I asked it to work independently, document its actions, avoid inventing facts, and seek permission before publishing, sending messages, or making commitments.
Then I hit enter.
Twenty-six seconds later, the results were in.
That was fast. But speed was only the beginning of the test.
Where AI Delivered—and Where It Fell Flat
1. News research: A promising start
Terra identified three news opportunities, explained why they mattered to my audience, and recommended when to publish them.
The output was structured and useful. It gave me a starting point for deciding which stories deserved attention.
But identifying a promising story isn't the same as establishing that every detail is correct. In journalism, verification remains non-negotiable.
2. Article writing: Fast, fluent, and frustrating
This was where the cracks started to show.
I asked for 800 words. Terra gave me 921.
That might sound like a minor issue, but following a word limit is part of following the brief. If I have to spend time cutting an article down before I can even begin editing it properly, the supposed efficiency starts to disappear.
The bigger problem was the writing itself.
It was dry. Repetitive. Full of polished-sounding phrases that said very little. The kind of prose that looks professional at a glance but starts to unravel when you read it closely.
Even the familiar AI-ism “that distinction matters” made an appearance.
As a journalist, I wouldn't publish it. I would have to rewrite substantial sections, sharpen the argument, remove the fluff, and inject a voice that sounded like an actual human being with something to say.
At that point, the 26-second turnaround didn't feel quite so impressive.
AI can produce a draft in seconds. Making that draft worth reading is another job entirely.
3. Interview preparation: Competent, not compelling
Terra generated ten interview questions for a technology executive. Most were usable, but one or two lacked the depth and originality I wanted.
That matters when you're interviewing someone who has answered the same predictable questions dozens of times.
A good interview question doesn't simply invite an explanation. It challenges assumptions, exposes a contradiction, or gets an executive to reveal something genuinely interesting.
AI can give you the framework. Finding the question that makes someone stop, think, and offer a revealing answer still takes judgment.
4. LinkedIn and audience growth: All the right parts, none of the spark
The LinkedIn post looked fine on the surface. It had the structure, the key points, and the familiar ingredients of a social media update.
What it lacked was a compelling opening.
There was no immediate reason to stop scrolling. No sharp insight. No line that made me want to read on.
The same problem can affect AI-generated newsletters and distribution plans: they can be technically complete without being particularly effective.
When everyone has access to tools that can produce competent content on demand, competence alone isn't much of a competitive advantage.
I would have needed to make substantial edits before publishing.
5. Business development: The strongest result
This was Terra's best performance.
It identified promising commercial opportunities from contacts and organizations already within my professional orbit. Unlike the writing tasks, the recommendations felt directly relevant to my existing work.
That made the output genuinely useful.
It also highlighted an important distinction: AI can be particularly effective when it has relevant context, clear criteria, and a concrete objective.
Finding a potential partner is one thing. Building a relationship, understanding what motivates them, and negotiating a deal are different challenges.
Still, this was the task where I could most clearly see AI saving me time.
Beyond Writing: Where AI Gets Interesting
Terra's capabilities extend beyond traditional knowledge work.
The model can also help with tasks such as:
Analyzing videos frame by frame.
Interpreting charts and technical drawings.
Completing forms and other computer-based workflows.
Working with visual material, including drawings that may be developed into 3D models.
I also tested a three-minute video by asking it to generate a lifelike portrait based on my appearance. The result was recognizable, although it included a minor limb distortion that needed correcting.
These are interesting capabilities, particularly for people whose work involves visual analysis and repetitive digital tasks.
But they reinforce the same lesson: the more specialized the output, the more important it is to check whether it's actually right.
The Real Test Isn't Speed. It's How Much Work You Get Back.
Here's the problem with judging AI by how quickly it responds.
A task that takes 26 seconds to generate can still take an hour to review, correct, fact-check, and rewrite.
A draft that sounds polished can still be unsuitable for publication. Ten interview questions can still produce a mediocre interview. A social post can tick every box and generate no meaningful engagement.
The time saved at the beginning may simply reappear at the editing stage.
That's why I think the most useful question isn't, “How fast can AI do this?”
It's “How much human work does it eliminate without compromising quality?”
To answer that, we need to measure more than response time. We need to look at accuracy, originality, usefulness, the amount of editing required, and whether the final result is genuinely better than what we could produce ourselves.
On some tasks, Terra delivered. On others, it gave me another layer of work to do.
And that distinction is critical for anyone evaluating AI at work.
Will AI Replace My Job? Not Yet.
I came away impressed by Terra's speed and its potential for handling computer-based tasks. Its business development recommendations were useful, and its visual capabilities point to some interesting possibilities.
But I wouldn't hand over my entire working day to an AI agent and assume everything would come back ready to use.
Journalism requires more than assembling facts into sentences. It requires judgment, skepticism, original thinking, a feel for the audience, and the ability to recognize when a technically correct answer misses the point.
The same is true of many knowledge-work roles. Producing an output is only one part of the job. Knowing what deserves to be produced—and whether it's good enough—is another.
AI will undoubtedly change how we work. In some cases, it will eliminate tedious tasks and free up time for higher-value work. In others, it will generate a convincing first draft that still needs considerable human intervention.
My takeaway from this experiment is simple: don't measure AI by how much work it produces. Measure it by how much work you no longer have to do.
Because 26 seconds is impressive.
Getting the job right is what counts.