Pro tip, use o1 and just talk to it, tell it the full task and tell it to make you a detailed SOP step by step for an operator agent to follow to complete this task. And use that as the prompt
@@gabrielsmith8767 Yep, when I woke up the morning after setting it up, everything was complete except for a few final questions that required personal information.
Has anyone tried operator chaining (like prompt chain)? So operators can colaborate, verify actions and correct them if necessary to reach the desired outcome?
If you point operator at operator you get a funny message. You CAN point it to chatgpt though, and amazingly you can point it to Gemini deep research as well (and presumably OpenAI deep research as well.
Could You connect Operator to your local machine through streaming it from the browser using for example Chrome remote desktop? And ask it to use software on the computer?
😅 OMG. Amazing. I didn’t think I spend the money but now with Deep Research…. We are on the verge of something big. I know that the knowledge cut off for the models has been updated recently too and even though that might not sound important, it is. We are getting closer….
Can you try using Operator for Test Automation: 1. Human creates a sheet with test data 2. Instruct Operator to use that data to complete a transaction in a SaaS application 3. Take screen shots upon completing each screen (or finding an error) and insert the screen shots in a new document to document the results. Thank you.
Great video! Why use Operator for web research when there’s Deep Research from Gemini that seems to be faster and generate way better research reports?
How about an operator that gets a short list of job descriptions and a CV. Changes the CV in google docs to fit the job description criteria and submit a job application?
I would assume being that I’ve see. Videos of people training it to do new things. But it’s using a basic Bing browser. So you could at least have it go to the actual coupon website, search for the company, and find a coupon. Then of course you would have to give it your login info (username and password).
Supposing you have N number of tasks running. Can you set Operator to send you a task message telling you that it's completed a task, or so you need to constantly check in with it. Use-case: i need quotes or buying prices for items from 50 vendors
Can you ask it to use deepseek to double check stuff and get more instructions? Also can you give it access to a powerful cloud machine using a service like shadow? I think this could be very powerful when you can run multiple operators with a supervisor role to ensure quality.
How does Operator compare to a WebUI Agent setup right now? In terms of accuracy I guess, not capability (ik operator is pretty constrained). I’m most interested in Presentation/Slides and Google Sheets use!! Thanks
Thanks for featuring useful ideas rather than booking a flight, or dinner reservations, or a massage, and for having good audio, all of which OpenAi seems to fail at with every new announcement. Their audio is some of the most amateurish on RUclips.
one curiosity I have.. if this takes off, in 1-2 years lots of folk using Operators to summarize data... why would anyone put together a website anymore (with new data). I mean, without human traffic there's no point in advertising on websites, without humans you're not getting community interactions anymore so there's no social benefit or prestige.. so what's the point? And then once that happens, what are these things going to scrape?
Don't underestimate the human desire to see things for themselves. I think human interactions will drop on some sites but I think you'll have mostly enthusiasts/specialists making and viewing sites.
@@teggerzz Books are (were) written so they could be sold for money through a publishing house, or so that the person's name would go out there into history. When they are skimmed and paraphrased there is zero money to be made from writing them, the author's name is removed so there is no recognition, and even the content is adjusted so the meaning of the original words is lost. So my question absolutely still stands, I am not sure how your statement actually addresses anything. And libraries are actually struggling to survive for that exact reason you mentioned, and largely survive through government grants. And the youngest generations largely do not use them anymore (aside from when taken there by their parents as kids).
@@Konarali That's possible, but I'm not sure there will be enough of those to keep things going using that model. We all know a lot of enthusiasts for some niche project that tried to create websites or pod-casts and then dropped off from lack of viewers. And this directly goes toward lack of viewers as data is scraped. And yes, we could say that people will seek 'humans for human contact'... but its already getting hard to know who the humans are, or if you're talking to the human or someone on their staff or a bot. So unless you know the enthusiast in-person, are you sure they're real?
I tried Gemini on google for my healthcare notes for my job. It works! Just not as personal nor smart as ChatGBT. It’s kinda just a little tool that’s there. But they seem to be integrating it in every google software.
Is that true for sure?? I’ve been wondering about stuff like this… government websites seem notoriously “guarded” …or something? I can’t quite figure it out. I’m guessing web devs just don’t like bots?? 🤷
That is the AI…? He is mostly showing the screenshots of each step the AI was completing. Which they only show the results as screenshots. Not a smooth video of a cursor scrolling or moving. Just straight points on the page. That’s how the AI works bc it sees a grid then selects that item in the grid. They are saving a lot of processing power by not currently being “pretty.”
Pro tip, use o1 and just talk to it, tell it the full task and tell it to make you a detailed SOP step by step for an operator agent to follow to complete this task. And use that as the prompt
ha, pro tip, ha, cause you need Pro to access it
@ yes pro is a no brainer
@ I mean, deepseek works and it’s free.
@alexatedw lemme enjoy the pun bro kek
Sorry for n00b question but what does SOP stand for? Thanks!
Best video on operator yet!! Congrats
"The future is here!" You look so genuinely happy saying that; I love it!
You got me so excited about operator! You’ve got me thinking about so many new use cases
and people are comparing Deepseek to ChatGPT which has this amazing tool
In the past video you teased that you're using Operator to research Operator Igor, that's crazy. You're nailing it.
I may or may not have had it complete my 6 hour traffic school course.
Man, that's a good one, lmao 😂😂😂
Bro did this work??? Can it work this long?
@@gabrielsmith8767 Yep, when I woke up the morning after setting it up, everything was complete except for a few final questions that required personal information.
MORE CONTENT ON THIS TOPIC PLEASE
Was looking forward to it, let's see! :)
Has anyone tried operator chaining (like prompt chain)? So operators can colaborate, verify actions and correct them if necessary to reach the desired outcome?
Following
I've done it using langchain
@@TheAlchemist1089 thanks! New to AI I will be checking that out
If you point operator at operator you get a funny message.
You CAN point it to chatgpt though, and amazingly you can point it to Gemini deep research as well (and presumably OpenAI deep research as well.
Thank you. This is awesome.
Could You connect Operator to your local machine through streaming it from the browser using for example Chrome remote desktop? And ask it to use software on the computer?
I asked about this one, is on the to do list lol
Thanks so much for your great video ❤❤
😅 OMG. Amazing. I didn’t think I spend the money but now with Deep Research…. We are on the verge of something big. I know that the knowledge cut off for the models has been updated recently too and even though that might not sound important, it is. We are getting closer….
Thank you! This is very powerful!!!
19:05 - great to see these use cases
Can it visit GPT operator and loop himself 2 times?
nope, operator is blocke bit you could use external apps to control it
Thanks, Good to know and that is expected:)
Can you try using Operator for Test Automation: 1. Human creates a sheet with test data 2. Instruct Operator to use that data to complete a transaction in a SaaS application 3. Take screen shots upon completing each screen (or finding an error) and insert the screen shots in a new document to document the results. Thank you.
Great video! Why use Operator for web research when there’s Deep Research from Gemini that seems to be faster and generate way better research reports?
You wouldn't
How about an operator that gets a short list of job descriptions and a CV. Changes the CV in google docs to fit the job description criteria and submit a job application?
Can operator analyze online videos? And can it read a computer’s file system to eg organize it
Can it go to a checkout page, then apply coupons from browser extensions ?
I would assume being that I’ve see. Videos of people training it to do new things. But it’s using a basic Bing browser. So you could at least have it go to the actual coupon website, search for the company, and find a coupon. Then of course you would have to give it your login info (username and password).
Wait… there is an extension tab on the top right corner on the Operator’s Browsing Screen.
@@gabrielsmith8767 Yep there is. That's why I got curious. No one has yet showed it's use case
Hope they release it to Pro users soon. I currently have a boring task I could really use it for. Not for $200 though.
What a beautiful zone. Thanks for taking us deep once again. Loved it.
I tried to get it to research AI tools and put it's findings into a spreadsheet and it got really confused about scrolling around the sheet.
Supposing you have N number of tasks running. Can you set Operator to send you a task message telling you that it's completed a task, or so you need to constantly check in with it. Use-case: i need quotes or buying prices for items from 50 vendors
Can you ask it to use deepseek to double check stuff and get more instructions?
Also can you give it access to a powerful cloud machine using a service like shadow?
I think this could be very powerful when you can run multiple operators with a supervisor role to ensure quality.
You would need a tool like n8n for this kind of thing
I'm waiting for when it's good enough to pop out my calculus 2 homework in my own handwriting😅
Is it possible to access this somehow from the EU or do I have to move to the US?
I just use a VPN and set the location to US
Manual QA tests of apps? :)
How does Operator compare to a WebUI Agent setup right now? In terms of accuracy I guess, not capability (ik operator is pretty constrained). I’m most interested in Presentation/Slides and Google Sheets use!! Thanks
Does operator work in Make? That would make it worth my money. P.S. Love what your group is doing.
Would love to know this too lol
Thanks for featuring useful ideas rather than booking a flight, or dinner reservations, or a massage, and for having good audio, all of which OpenAi seems to fail at with every new announcement. Their audio is some of the most amateurish on RUclips.
Regarding the audio thing, I actually kind of like it, gives it much more of a live vibe, much less produced if you know what I mean
Any issues using operator over vpn as its US only?
Can it access ChatGPT o1 model in web and ask it different prompts?
one curiosity I have.. if this takes off, in 1-2 years lots of folk using Operators to summarize data... why would anyone put together a website anymore (with new data). I mean, without human traffic there's no point in advertising on websites, without humans you're not getting community interactions anymore so there's no social benefit or prestige.. so what's the point? And then once that happens, what are these things going to scrape?
Don't underestimate the human desire to see things for themselves.
I think human interactions will drop on some sites but I think you'll have mostly enthusiasts/specialists making and viewing sites.
“Why write books or keep libraries if all of it is on the internet”?
Don’t get old behind your time. Use your brain.
@@teggerzz Books are (were) written so they could be sold for money through a publishing house, or so that the person's name would go out there into history. When they are skimmed and paraphrased there is zero money to be made from writing them, the author's name is removed so there is no recognition, and even the content is adjusted so the meaning of the original words is lost. So my question absolutely still stands, I am not sure how your statement actually addresses anything.
And libraries are actually struggling to survive for that exact reason you mentioned, and largely survive through government grants. And the youngest generations largely do not use them anymore (aside from when taken there by their parents as kids).
@@Konarali That's possible, but I'm not sure there will be enough of those to keep things going using that model. We all know a lot of enthusiasts for some niche project that tried to create websites or pod-casts and then dropped off from lack of viewers. And this directly goes toward lack of viewers as data is scraped. And yes, we could say that people will seek 'humans for human contact'... but its already getting hard to know who the humans are, or if you're talking to the human or someone on their staff or a bot. So unless you know the enthusiast in-person, are you sure they're real?
@@hypnokitten6450 I guess you won't but this could lead a push for verified human sites. The outcome of that is those sites are scraped.
is there a free tier of The AI Advantage Community? it's super expensive for me
great video. does anyone know if Gemni can already do these tasks without the need to join GPT $200 per month?
I tried Gemini on google for my healthcare notes for my job. It works! Just not as personal nor smart as ChatGBT. It’s kinda just a little tool that’s there. But they seem to be integrating it in every google software.
perfect
So basically, Operator is a more advanced version of Tasks 🤔
Careful, Midjourney will ban accounts for using operator on their service.
Is that true for sure?? I’ve been wondering about stuff like this… government websites seem notoriously “guarded” …or something? I can’t quite figure it out. I’m guessing web devs just don’t like bots?? 🤷
I’m glad ai also struggles with the GitHub UI 😂
If we can get an open sorce operator that we can run locally imma have to build a custom PC
Can Operator pilot games?
Rn it only works on browser pages.
@@gabrielsmith8767 yup. there are a lot of browser games
Operator - DJ Koze's Disco Edit
ruclips.net/video/uvzJ-3yCto0/видео.html
Wonderful, I pay 200$ per month, want to use operator - not available in your region (Romania)
Annoying I know. Use a VPN and it will work!
We are behind EU "paywall"...
Operator can Login in Paid sites with my credentials if I ask?
100 % yeah
Good topic but STOP MOVING THE CURSOR like a 5 year old . Do you have ADHD . I lose interest when you constantly scroll and move the mouse.
That is the AI…? He is mostly showing the screenshots of each step the AI was completing. Which they only show the results as screenshots. Not a smooth video of a cursor scrolling or moving. Just straight points on the page. That’s how the AI works bc it sees a grid then selects that item in the grid. They are saving a lot of processing power by not currently being “pretty.”