Episode 124: Vibe Working: Running Parallel Agent Teams to Replace Busywork

When your agent is finished and waiting for you, you get stressed because you didn't give them a job, so you want to make sure every night or weekend when you're leaving, you keep your agents busy so they keep working.

There is that agent stress like your agents are waiting for you, so whoever I'm talking to, they're seeing that Cloud Code is handling everything and getting the job done, so we are not writing code anymore.

Hi, my name is Demetri Panici and I'm a content creator, agency owner, and AI enthusiast. You're listening to the AI Agents podcast brought to you by Jotform and featuring our CEO and founder, Aytekin Tank. This is the show where artificial intelligence meets innovation, productivity, and the tools shaping the future of work. Enjoy the show.

Hello and welcome back to another episode of the AI Agents Podcast. Today we're going to be talking about vibe working and how to manage your AI agents, which is something a lot of people are discussing right now.

Vibe working in 2026 is something we've both been trying to do, and we were even fiddling with some stuff before the call to help you be more in touch with what we're doing. I'm excited to show the stuff we're doing here.

Why coding has been a thing since last year, but recently especially with Cloud Code's improvements, coding became so good that whoever I talk to, including friends with startups, they're seeing that we are not writing code anymore.

One of my friends has a team of AI agents that he's constantly running to write code, and when one agent finishes the job, he assigns the agent to do something else. There's even a new stress called agent stress, which is when your agent is finished and waiting for you, and you get stressed because you didn't give them a job.

So you want to make sure every night or weekend when you're leaving, you keep your agents busy so they keep working. You can even run agents in parallel now, and we're going to showcase that today.

Whoever I'm talking to sees that Cloud Code is handling everything and getting the job done, so we are not writing code anymore. At Jotform, we also use Cursor and Cloud Code, and some people develop internal tools just by wipe coding.

Usually when we develop our product, AI is just a helper and not taking over the job yet. This has been happening recently in the last couple of months, and there was even an article about something big happening.

The article was called something big is happening, showing a person holding a piece of paper. What's happening is that all these tools developed for wipe coding can actually do other things, like using the browser or any software such as spreadsheets or document editors.

So if you are doing office work, AI can actually do those things. We showcased Cover recently, which was amazing. Last week we covered Open Clove, which constantly changes its name. Open Clove was the last name I believe.

It's amazing how fast and crazy it is, doing everything quickly, making you always watch what it is doing because it's kind of scary. Co-work is much slower, very safe, sometimes too slow and too safe.

The demo I will do today is on Cover because I want to showcase parallel agent teams and how agents can run in parallel so you can do the job much more quickly and get things done.

You can have different agent specialties, like a designer, developer, researcher, competitor researcher, and all these agents can work together communicating with each other. This is Cloud's agent teams feature, and it's really cool stuff.

The term wipe working was coined on a LinkedIn post, which I think you're going to share.

The post starts by saying today is February 6th, and I need you to brace yourselves because this was the week AI stopped being a conversation and became an earthquake. 285.85 billion dollars vanished from the global software industry in a day, not because of recession or war, but because of a plugin.

Anthropic launched new capabilities for AI agent co-work tools that can review contracts, NDAs, manage compliance workflows, and draft legal documents. Wall Street's reaction was immediate and brutal, with Thomson Reuters dropping 18%, the worst day in the company's history.

Relics, the parent company of LexisNexis, fell 14%, the worst since 1988. LegalZoom plunged nearly 20%. The head of equity research at LPL Financial asked why continue paying for software when developing these systems is becoming faster and more efficient with AI.

It's not a prediction, it's a eulogy. Scott White, Anthropic's head of enterprise, said in an interview that we are now transitioning into vibe working, where knowledge workers hand off entire tasks to AI agents using natural language for financial analysis, legal research, documentation generation, end to end.

OpenAI's CEO dropped the most important question of the year: does software still matter? That's a good question.

The question is what will happen to software businesses. It will be really challenging for many SaaS businesses, probably including us, because you can do lots of things with this software.

But it's also good because it will push us towards being more valuable for our users. Good software doesn't just do stuff for you; it helps you, tells you the best way to do something, and guides you.

It's not just about building forms, but with our new AI stuff, it's about making sure whatever you're trying to do, we help you build the right workflows, emails, and analyze data properly.

If you have a software business that's just one prompt, it can be easily replaced. But we are still far from building reliable software with a single prompt. It takes a long time.

Most of the time, AI-created apps are good prototypes, but advancing them gets harder because you have to think about compliance, security, integration with other software, and many other things that are not easy.

It's interesting times. The software business will also change. Software products will become less rigid tools and more like tools that understand what you're trying to do and help you accomplish those things with AI.

Sales companies are also adding AI to their products. We've been doing some cool stuff with Jotform that we can cover in another episode.

The article 'Something Big is Happening' by Matt Schumer went viral on Twitter and LinkedIn. Many people don't see the value in return for their business and think about it individually.

I want to acknowledge how fast things are moving. In 2022, AI could do basic math reliably; by 2023, it could pass the bar exam; by 2024, it could write working software and explain graduate-level science.

By late 2025, some of the best engineers said they handed over most of their coding work to AI. On February 5th, new models arrived that made everything before them feel like a different era.

I know you worked with sub-agents and parallel agents inside Claude Co-work and Cloud Code. I'm doing similar stuff, and I'm curious what you think is best moving forward for business owners.

There's a lot of work I try to offload, especially in the last six weeks with Opus 4.5 and 4.6. These models are game-changing for me because I can open my computer and just do stuff.

I noticed it takes a lot of work to get it going in the right direction. You need to know what you're doing with tools like Open Claw to avoid infinite loops and security issues.

As an individual, this can be really powerful if you spend the time on it, but as a business owner, it can be frustrating to find returns with the cost where they're at.

Many of these tools are heavily subsidized. For example, the Claude Co-work max $200 plan is worth $20,000 a month in tokens if used to its fullest extent.

This is an urgent call to learn how to utilize these tools to set up workflows in the short term because in the long term, it will cost more to do the same things.

Frontier models like Claude and OpenAI will keep getting better, but local open-source models are also improving and getting easier and cheaper to run locally.

There will be a point when subsidies end, and investors will want to recoup their investment. For now, we are being subsidized as a nerd group paying attention to this stuff.

There's huge competition among Cloud, ChatGPT, and Gemini, each gaining momentum with their own advantages. Cloud Code is popular with developers, Gemini has a huge Google user base, and ChatGPT has billions of users.

The winner is not clear and won't be for a long time because the market is evolving quickly with new products and improvements.

It's a great time for us because we can use these products to do lots of great stuff. For example, simple AI-generated motion graphics can now be made quickly with Cloud Code and Cloud Co-work.

I spent a day putting together visuals and video editing engines with AI that would have cost thousands in tokens but instead cost a day of the plan each time, which is incredible.

I can finish a meeting, have a paid ads proposal put together, align on next steps, and tell Cloud Code to analyze the transcript and do the work, only checking in to add prompts or connect tools.

OpenAI 4.6 has dropped and is continuously improving computer understanding, UI/UX, thinking, and spreadsheet capabilities, making vibe working more possible throughout the year.

It would take a video editor about three hours to make one motion graphic like this, but AI made 500 iterations in about eight hours while I was doing other stuff.

Our team used Freepik for motion graphics on Jotform's 20-year anniversary videos, and they were impressed with the quality and ease of use.

Creative AI stuff is underrated. Some media communities poo-poo AI, but there's opportunity in every realm, including motion graphics, office work, and paid ads.

An AI-generated Spider-Man trailer showed a future where Spider-Man is sad and Mary Jane is in a wheelchair. Some people love it as the future of cinema, while others reject it as cringey nostalgia.

Being human is important, but there's opportunity in media and other fields to use AI to take ideas and automate tasks with tools like Cloud Code and Cloud Co-work.

Figure out how to get the most out of these plans in the short term because Anthropic removed Open Claw access yesterday, possibly due to OpenAI buying Open Claw.

Open Claw is better than Co-work but unsafe and dangerous. OpenAI buying it made Anthropic unhappy because they are competitors.

There are pricing discrepancies and privacy concerns that may be addressed soon. People using Open Claw with the $200 max plan got $20,000 worth of value, which is unusual.

Anthropic removed access to Open Claw and their own agent SDK's OOTH token authorization for flawed plans, which is fascinating.

You can still get a lot out of Open Claw by building it through the Cloud Code terminal, but it requires extra technical understanding. Alternatively, you can just use Co-work if you want less technical setup.

I want to show how to use parallel agents in Co-work. You can build agents with different skills, like screenshot taker or web page builder, and have them work together.

For example, research 10 sales products powered by AI, launch parallel agents to research each product, and create a slideshow. This saves time and lets you leave the work running.

Each agent does deep research on its assigned product, using the browser to visit websites and gather information, then contributes to a presentation.

When working on a product feature, I can ask Co-work to find 100 different co-pilots or similar products, take screenshots, create reports, and compare features quickly, which would take a product manager a month.

The AI is still working on this, doing web searches and creating a PowerPoint with pages from different agents. This is a simple example of what can be done.

If you have a cross-functional team of agents like designer, web developer, and product manager, you can train each agent with specific skills and assign projects to the team.

You can give projects to these teams and have them complete work overnight, delivering products or reports by morning. This is amazing and shows the direction of AI work.

Agent teams avoid context mixing and token limits by having each agent focus on their own skills, making the work more efficient and reliable.

I will try to use agent teams for another demo in the future. I'm also working on other projects to showcase soon.

I want to show you an example of my Open Claw called Jerry. It's a mission control I built to make Open Claw easier to understand and use.

Because of token issues, I used Cloud Code to build this out. Cron jobs and workflows are stored on your computer, with some data sent to the server.

Jerry Baker is my Claude Code setup that I use for paid ads. I'm building an AI-powered ads dashboard to analyze and optimize ad performance across platforms without manual work.

Claude Code and Claude Co-work are similar, but Claude Code lets you build stuff with coding, while Co-work is limited to non-coding capabilities.

The motion graphic I showed earlier was made with Cloud Code by prompting it to find how to create it on YouTube, then implementing it with Remotion and terminal commands.

If you enjoy Claude Co-work, try playing around with Claude Code because it can build dashboards and other tools by itself without you learning to code.

I will show a demo of adding to the dashboard. Jerry is busy doing everything and you can communicate with it via email, Telegram, WhatsApp, and other tools.

You can send commands through Telegram, but it's best to build cron jobs and plans in Cloud Code because it's cheaper, then have Open Claw execute them on a recurring basis.

I gave Jerry a large prompt to communicate between different parts, and it's working on making a heat map and other features, though sometimes it can be finicky and require refreshing.

Open Claw has a dashboard to understand its backend and chat with it directly, but I prefer Telegram for ease of use on the go.

Demos sometimes break during the show, which is frustrating, but I can fix issues quickly by copying and pasting into Cloud Code.

I'm working on multiple projects and can't show everything now, but vibe coding and vibe working are very valuable right now, letting you automate many daily tasks.

Think about the different things you do daily and how they can be automated, like analyzing email, responding to Slack messages, or managing to-do lists by typing in chat.

Vibe working makes daily and work tasks easier by automating them with AI agents, letting you delegate research or reports overnight and get useful results the next day.

Agent teams working in products will speed up work significantly, letting you give outcomes and get jobs done in minutes with huge potential and momentum.

Check out the article 'Something Big is Happening' and everything we're doing. We're available everywhere and appreciate your support. If you liked this video, hit like and subscribe. We'll talk more about this in the future. See you next time.