DEV Community: Vivek Shetye

I stopped asking one AI chatbot for startup ideas. 🤖 Instead, I built a 4-agent Startup Intelligence team with Hermes Agent to research markets, challenge weak evidence & design validation experiments. Here’s how it works 👇

Vivek Shetye — Tue, 11 Aug 2026 18:58:37 +0000

Hermes Agent Challenge Submission: Build With Hermes Agent

Vivek Shetye

Aug 11

I Built a Team of AI Agents to Find Startup Opportunities

#showdev #ai #hermesagentchallenge #startup

7 min read

I Built a Team of AI Agents to Find Startup Opportunities

Vivek Shetye — Tue, 11 Aug 2026 18:37:09 +0000

Most people use AI for startup research like this:

“Give me 10 promising AI startup ideas.”

A few seconds later, you get a polished list.

The problem?

You have almost no idea which conclusions are backed by evidence, which are assumptions, and which are simply the model confidently connecting dots.

So I tried something different.

Instead of asking one AI agent to find startup ideas, I built a small Startup Intelligence team using Hermes Agent.

The system uses four specialized AI agents that research markets, investigate competitors, audit evidence, challenge each other’s conclusions, and ultimately rank promising B2B AI SaaS opportunities.

And rather than producing another Markdown document full of ideas, the workflow produces structured research containing:

Market opportunity scores
Companies and competitors
Customer pain and unmet needs
Evidence-backed claims
Source URLs and supporting passages
AI advantages and workflows
Low-cost validation experiments

Here’s how the system works.

🎥 Full video walkthrough

The Problem With Asking One AI Agent to Find Startup Ideas

Startup research looks easy until you actually need to decide where to spend your time and money.

A few signals can be surprisingly misleading.

💰 Funding can look like customer demand.

A market receiving hundreds of millions in venture capital doesn’t necessarily mean customers are willing to pay for another product.

📈 Growth claims can look like market validation.

Especially when the numbers come directly from vendors.

🏢 Customer logos can look like retention.

A logo doesn’t tell you how much the customer pays, how heavily they use the product, or whether they’ll renew.

⚔️ A long competitor list can make a market look saturated.

But those companies may target completely different buyers, workflows, or budgets.

Generic AI research tends to compress all these signals into something like:

“This is a rapidly growing market with strong demand and significant opportunity.”

That sounds convincing.

But as a founder, it doesn’t tell me what I actually need to know:

Is this opportunity strong enough to investigate further?

So instead of optimizing the system for generating ideas, I optimized it for reducing uncertainty.

Building an AI Startup Intelligence Team

I created four separate Hermes Agent profiles.

Each agent has:

its own role
its own instructions
its own research skill
its own memory
a clearly defined responsibility

The architecture looks roughly like this:

                     ┌─────────────────────┐
                     │   Startup Director  │
                     └──────────┬──────────┘
                                │
                    Defines scope + rubric
                                │
                ┌───────────────┴───────────────┐
                ▼                               ▼
      ┌───────────────────┐          ┌─────────────────────┐
      │ Market Researcher │          │ Competition &       │
      │                   │          │ Signals Analyst     │
      └─────────┬─────────┘          └──────────┬──────────┘
                │                               │
                └───────────────┬───────────────┘
                                ▼
                     ┌─────────────────────┐
                     │   Skeptic Editor    │
                     └──────────┬──────────┘
                                ▼
                      Evidence-backed
                     research package

The important part isn’t simply having four agents.

It’s giving them different judgment lenses.

Let’s look at each one.

1. Startup Director — The Orchestrator 🧭

The Startup Director acts like the project lead.

Its first job isn’t browsing the web.

Instead, it converts an ambiguous founder question into a bounded research problem.

For example:

Find five promising B2B AI SaaS startup markets in North America with meaningful evidence of customer demand and new company formation between January 2025 and August 2026.

Before delegating research, the Director creates two important artifacts:

brief.md
rubric.md

The brief defines things such as:

geography
buyer
industries
time window
company scope
exclusions
deliverables
known unknowns

The rubric defines:

source tiers
scoring dimensions
claim statuses
evidence requirements
counter-evidence requirements
final output structure

This matters because otherwise different agents can quietly interpret the same research question differently.

One might optimize for funding.

Another might optimize for TAM.

Another might optimize for how many startups exist.

The rubric establishes the rules before the evidence is collected.

2. Market Researcher — Is the Pain Real? 🔎

The Market Researcher investigates whether there is meaningful evidence that customers actually have the problem.

It looks for signals including:

buyer pain
willingness to pay
customer demand
company formation
funding
adoption
traction
hiring signals

But simply finding a claim isn’t enough.

The research needs to preserve where that claim came from.

That distinction becomes extremely important later when another agent audits the research.

3. Competition & Signals Analyst — What Already Exists? 🧩

The second researcher approaches the same markets from another direction.

Instead of primarily asking:

“Is there demand?”

It asks:

“How is this problem being solved today?”

That means researching:

startups
incumbents
internal tools
agencies
spreadsheets
existing software
adjacent products

It also investigates whether market attention is translating into something stronger:

adoption, payment, retention, or recurring usage.

This distinction matters.

A market can be extremely popular on X, LinkedIn, Product Hunt, or Hacker News while having surprisingly little evidence that businesses are paying to solve the problem.

The Market Researcher and Competition Analyst therefore examine overlapping markets but with different objectives.

4. Skeptic Editor — Try to Break the Thesis 🛡️

This might be the most important agent in the entire workflow.

Most AI pipelines optimize for generating an answer.

The Skeptic Editor optimizes for finding reasons that answer might be wrong.

It doesn’t simply read the summaries produced by the other agents.

It examines the underlying:

research files
claims ledger
source URLs
supporting passages
claim classifications
counter-evidence

Then it asks questions like:

Does this citation actually support the claim?
Is this vendor-reported or independently verified?
Does the evidence cover the correct market?
Does it fall inside the requested time window?
Are we confusing funding with customer demand?
Is a company-reported metric being presented as independent evidence?
What evidence contradicts the recommendation?

If a source supports only a weaker statement, the claim gets narrowed.

If the evidence quality is poor, confidence gets downgraded.

And if there isn’t enough evidence?

The answer can simply become:

UNKNOWN

That’s a feature, not a failure.

An AI research system should be able to admit uncertainty.

Orchestrating the Agents With Hermes Kanban

There was another architectural problem.

Separate Hermes profiles don’t share conversation history.

That means I needed a durable coordination mechanism.

This is where Hermes Kanban becomes useful.

The Startup Director creates two independent research tasks:

Market Research
Competition & Signals

Because neither depends on the other, they can run in parallel.

The Director then creates another task:

Skeptical Review

But this task has both research tasks as parents.

Conceptually:

             Startup Director
                    │
           ┌────────┴────────┐
           ▼                 ▼
     Market Research    Competition
           │                 │
           └────────┬────────┘
                    │
             BOTH COMPLETE
                    │
                    ▼
             Skeptic Editor

The Skeptic Editor therefore doesn’t start synthesizing conclusions while half the research is still missing.

This creates an actual dependency graph rather than simply launching several agents and hoping they coordinate.

What Did I Ask the Agents to Research?

For this experiment, I asked the system to compare five specific B2B AI SaaS startup markets in North America.

The industry scope covered:

🏥 Healthcare & Life Sciences

💻 Software Development

🎧 Customer Support & Contact Centers

📊 Finance & Accounting

🛍️ Retail

I also explicitly excluded:

generic wrappers around foundation-model APIs
consumer applications
AI infrastructure
chips
data centers
model providers
undifferentiated chatbots and copilots

The goal wasn’t:

“Which AI industries are growing?”

That’s too broad to be useful.

The agents instead compare specific combinations of:

Buyer + Pain + Workflow + AI Advantage

That’s much closer to the level at which a founder can actually validate an opportunity.

Every Important Claim Needs Evidence 📚

One requirement fundamentally changed the quality of the output:

Every important market claim needed to be backed by a source.

The claims dataset preserves fields such as:

claim
status
company / market
claim type
source URL
source title
publication date
retrieval date
source tier
supporting passage

This makes the research auditable.

Instead of seeing:

“Companies are increasingly adopting AI for this workflow.”

You can inspect the exact evidence that caused the system to make that statement.

And that means another agent—or a human—can challenge it.

The Final Output Isn’t Just a Report

The Skeptic Editor produces several structured CSV files.

markets.csv
experiments.csv
competitors.csv
companies.csv
claims.csv

This makes the results much more useful than one huge research document.

markets.csv

This contains the ranked opportunities.

For each market, the dataset can include:

market definition
opportunity score
target buyer
workflow
recommendation
companies
customer complaints
unmet needs
AI advantage
evidence quality

Now I can compare opportunities instead of reading five unrelated research reports.

companies.csv

This captures the companies discovered during research.

For example:

Company
Product
Buyer
Market
Differentiation

This helps answer:

Who is already attacking this problem, and how?

competitors.csv

Competition isn’t limited to startups.

The real alternative could be:

an incumbent SaaS platform
an internal operations team
an agency
a spreadsheet
a manual workflow

Understanding the status quo is often more useful than counting startups.

claims.csv

This is effectively the evidence ledger.

Every important conclusion can point back to the evidence supporting it.

That creates a useful chain:

Recommendation
      ↓
Market finding
      ↓
Claim
      ↓
Source
      ↓
Supporting evidence

This makes the final recommendation much harder to fake with confident language.

The Most Useful Output Might Be experiments.csv 🧪

Finding an interesting market still doesn’t mean you should build a product.

So every promising opportunity gets converted into a small validation experiment.

The experiments can contain:

Hypothesis
Target buyer
Procedure
Channel
Sample size
Cost / effort
Success threshold
Failure threshold
Decision rule
Evidence required

This changes the final question.

Instead of:

“Should I build this startup?”

The system asks:

“What’s the cheapest experiment that could prove this opportunity weaker or stronger?”

That’s a much better founder question.

AI Agents Should Reduce Uncertainty, Not Manufacture Confidence

After building this workflow, this was my biggest takeaway.

The value of multi-agent systems isn’t simply that you can run more AI agents.

Four agents producing four versions of the same answer isn’t necessarily better than one.

The advantage appears when agents have:

different responsibilities + independent evidence + explicit dependencies + durable state + adversarial review.

Then you’re building something closer to a research process.

The final system doesn’t claim:

“I found the perfect startup.”

Instead, it tells you:

what appears promising
what evidence supports that conclusion
what evidence contradicts it
what remains unknown
what you should test next

That distinction matters.

Because when you’re deciding where to spend months building a product, a confident answer isn’t enough.

You need a process that’s harder to fool.

🚀 Built an AI video production workflow with Hermes Agent and Remotion. The agent researches a topic, writes the script, generates visuals, creates motion graphics, adds AI voiceover, music, sound effects, animated captions, and renders the final video

Vivek Shetye — Tue, 04 Aug 2026 15:30:59 +0000

Hermes Agent Challenge Submission: Write About Hermes Agent

Vivek Shetye

Aug 4

🚀 Build an End-to-End AI Video Production Pipeline with Hermes Agent and Remotion

#ai #hermesagentchallenge #agents #automation

3 min read

🚀 Build an End-to-End AI Video Production Pipeline with Hermes Agent and Remotion

Vivek Shetye — Tue, 04 Aug 2026 15:04:33 +0000

Creating high-quality short-form videos traditionally requires switching between multiple tools for research, scripting, design, editing, voiceovers, captions, and rendering.

But what if an AI agent could orchestrate that entire workflow for you?

In this tutorial, I demonstrate how to build a complete AI-powered video production pipeline using Hermes Agent and Remotion. From discovering a viral topic to rendering a polished vertical video that’s ready for YouTube Shorts, Instagram Reels, or TikTok.

🎬 What You’ll Build

By the end of this tutorial, you’ll have an automated workflow that can:

🔍 Research trending topics from the web
📚 Collect sources and supporting references
✍️ Draft an engaging video script
🎨 Generate a detailed production plan
🖼️ Create AI-generated visual assets
🎥 Produce motion graphics with Remotion
🎙️ Generate natural AI voiceovers
🎵 Add background music and sound effects
💬 Create animated word-level captions using Whisper timestamps
📱 Render a production-ready vertical video

Instead of spending hours jumping between different applications, you describe what you want in natural language, and the agent coordinates the entire production process.

🎥 Full video walkthrough

🤖 Why Hermes Agent?

Hermes Agent isn’t just another coding assistant.

It can execute multi-step workflows that involve research, planning, asset generation, coding, media production, and iterative improvements.

For this project, Hermes Agent acts as an AI video producer by:

researching content ideas
writing scripts
planning visuals
generating production assets
building Remotion components
reviewing generated scenes
making improvements automatically

This allows creators to focus on storytelling instead of repetitive editing tasks.

⚡ Why Remotion?

Remotion lets developers create videos using React instead of traditional timeline-based editors.

That means your videos become:

version controlled
reusable
programmable
data-driven
easy to automate

When paired with Hermes Agent, Remotion becomes a powerful foundation for AI-generated motion graphics and automated video production.

🛠 The Workflow

The tutorial walks through the complete pipeline.

1️⃣ Research a Viral Topic

The first prompt asks Hermes Agent to research interesting recent discoveries about the universe.

Rather than returning a single answer, the agent:

searches the web
evaluates multiple sources
creates a structured research document
recommends the strongest topic

For the demo, I selected:

James Webb Space Telescope discovering methane on an interstellar comet.

2️⃣ Generate a Production Script

Hermes Agent converts the research into a complete production script containing:

narration
storyboard
Remotion cues
production notes
scene descriptions

Instead of just writing text, it prepares everything needed for video production.

3️⃣ Build a Scene-by-Scene Production Plan

Next, the agent generates a detailed production specification including:

🎨 color palette

🎞 animation style

📐 canvas resolution

⏱ frame rate

🎬 editorial direction

📋 scene breakdowns

This document becomes the blueprint for the entire video.

4️⃣ Generate Visual Assets

The workflow automatically creates AI-generated artwork for every scene.

Hermes supports multiple image generation providers, allowing you to use the model that best fits your workflow.

5️⃣ Generate the Motion Graphics Video

This is where Remotion shines.

Hermes uses the installed Remotion skills to:

generate React components
animate scenes
compose the timeline
organize assets
build the complete video

Even better, it reviews rendered frames against the production plan and iteratively improves scenes when needed.

6️⃣ Add AI Voiceovers

Once visuals are complete, Hermes generates narration using a text-to-speech model.

You can choose between:

local models
cloud providers
premium voices

depending on your setup.

7️⃣ Add Music and Sound Effects

The workflow also enriches the video with:

🎵 background music

🔊 sound effects

This makes the final result feel much more polished without requiring manual editing.

8️⃣ Generate Animated Captions

One of my favorite parts of the workflow.

Hermes uses Whisper to generate accurate word-level transcript with timestamps and creates captions that:

✨ highlight the currently spoken word

📈 animate smoothly

🎯 optimize readability

📱 are designed specifically for Shorts and Reels

The result looks significantly better than static subtitles.

💡 Why This Workflow Matters

AI video generation is evolving rapidly.

Many tools can generate clips.

Far fewer can manage an entire production pipeline.

This workflow combines:

AI research
scripting
planning
image generation
motion graphics
voice synthesis
caption generation
rendering

inside one cohesive workflow.

For developers and technical creators, this approach offers much greater flexibility than traditional video editors.

🎯 Who Is This For?

This workflow is ideal if you’re building:

educational YouTube channels
AI explainers
SaaS product demos
marketing videos
startup launch videos
developer tutorials
technical animations
social media content
automated content pipelines

💬 What Do You Think?

Would you trust an AI agent to produce your videos?

Or do you still prefer editing everything manually?

I’d love to hear how you’re using AI for content creation.

🚀 I Combined 3 AI Models Using Hermes Agent’s Mixture of Agents! Here’s What Happened

Vivek Shetye — Fri, 17 Jul 2026 19:45:56 +0000

We spend a lot of time asking one question:

Which AI model is the smartest?

GPT? Claude? DeepSeek? MiniMax?

But after experimenting with Hermes Agent’s Mixture of Agents (MoA), I started wondering whether that’s the wrong question.

What if, instead of choosing one AI model, you could make several models analyze the same problem and let another model combine their best ideas?

In other words:

3 AI Models → 1 Smarter Decision → Better Results

That’s the idea behind Mixture of Agents.

And I decided to put it to the test.

🤝 What Is Mixture of Agents?

Most AI agents work with a single model.

You send a prompt.

The model reasons about it, uses tools, and gives you an answer.

Mixture of Agents introduces another layer.

Your prompt is first sent to multiple reference models in parallel.

Each model independently analyzes the same problem and provides its perspective.

Their responses are then given to an aggregator model, which reviews the different approaches, combines the strongest ideas, and decides what to do next.

The reference models act like advisors.

The aggregator is the decision-maker.

Only the aggregator can:

📂 Read files
🛠️ Execute tools
💻 Run terminal commands
✍️ Modify code
✅ Produce the final response

So from your perspective, you’re still working with one AI agent.

Behind the scenes, however, that agent is consulting multiple AI models before making important decisions.

🏢 Think of It Like Asking Multiple Senior Engineers

Imagine you’re making an important architecture decision.

Would you rather ask one senior engineer?

Or have three senior engineers independently review the problem first?

One might immediately spot a security issue.

Another might question whether the architecture will scale.

The third might find a much simpler implementation.

None of them necessarily has the complete answer.

But a decision-maker who can see all three perspectives has more information before choosing a direction.

LLMs work in a similar way.

Different models have different strengths, weaknesses, biases, and reasoning patterns.

The goal of Mixture of Agents isn’t simply to add more models.

It’s to give the final model access to diverse perspectives before it commits to a solution.

🎥 Full video walkthrough

⚙️ How the Architecture Works

The basic workflow looks like this:

                  YOUR PROMPT
                       │
          ┌────────────┼────────────┐
          ▼            ▼            ▼
      AI Model     AI Model     AI Model
      Advisor      Advisor      Advisor
          │            │            │
          └────────────┼────────────┘
                       ▼
               Aggregator Model
                       │
                 Makes Decision
                       │
                       ▼
             Tool Calls + Final Answer

The important distinction is this:

Reference models advise. The aggregator acts.

The advisors don’t modify your project or execute tools.

They analyze the problem independently and provide additional perspectives that the aggregator can use when deciding what to do.

🧪 Putting 3 AI Models to Work

For my experiment, I configured:

⚡ DeepSeek V4 Flash — Reference Model
🚀 MiniMax M2.7 — Reference Model
🧠 GPT-5.4 — Aggregator Model

I then gave the agent a practical task:

Build a Kanban board for a solo YouTube creator.

When I submitted the prompt using /moa, Hermes sent the same problem to both reference models in parallel.

Each analyzed the task independently.

Once their responses were ready, the aggregator received:

My original prompt + DeepSeek’s perspective + MiniMax’s perspective

It could then synthesize those ideas and decide how to approach the actual implementation.

The result?

After roughly 14 minutes, the agent had built a functional single-page Kanban application with:

📋 Seven creator-focused workflow columns
📝 Editable task cards
🏷️ Tags and priorities
📅 Due dates
☑️ Checklists
🔍 Search
🎛️ Filtering
🖱️ Drag-and-drop cards
📊 Workflow statistics
💾 Local browser persistence

All from a single initial prompt.

The interesting part wasn’t just the finished application.

It was being able to see how multiple models contributed perspectives before the acting model committed to an approach.

👀 Watching Different AI Models Approach the Same Problem

One of the most interesting parts of Hermes Agent’s implementation is the visibility into what the reference models are doing.

While the task is running, you can see each reference model independently analyzing the problem.

This makes it possible to observe:

🧠 What different models notice
🔍 Details one model catches that another misses
💡 Different approaches to the same problem
🤝 How those perspectives inform the aggregator

It’s a useful reminder that different LLMs don’t always approach problems in exactly the same way.

And that’s where the potential value of Mixture of Agents comes from.

You don’t necessarily need three models to agree.

Sometimes the disagreement is the useful part.

🎛️ One Configuration That Can Make a Big Difference

One setting worth understanding is:

reference_max_tokens

This controls how much output each reference model can generate for the aggregator.

Giving every advisor thousands of tokens isn’t necessarily useful.

The aggregator often needs the key insights, not another complete solution.

Lower limits can therefore mean:

⚡ Faster responses
💰 Lower token usage
🎯 More focused advice
📉 Less unnecessary context for the aggregator

Hermes recommends 600 tokens as a practical default for concise reference responses, though the ideal value will depend on your task.

🧩 Different AI Teams for Different Problems

Another feature I found useful is the ability to create multiple Mixture of Agents presets.

That means you don’t need one universal AI team.

You could have:

💻 Coding

Models selected specifically for software engineering.

🔬 Research

Models with different strengths in analysis and long-context reasoning.

🏗️ Architecture

Models that provide complementary perspectives on system design.

📈 Financial Analysis

A completely different combination optimized for analytical tasks.

Instead of constantly asking:

“What’s the best AI model?”

The more interesting question becomes:

“What’s the best combination of models for this particular problem?”

🎯 When Does Mixture of Agents Actually Help?

I wouldn’t enable Mixture of Agents for every prompt.

If you’re asking:

What’s 15 × 27?

Three AI models aren’t going to make the answer three times better.

The additional reasoning becomes valuable when there are multiple valid approaches to a problem.

Some good candidates include:

🏗️ Software architecture
🔍 Code reviews
🐞 Complex debugging
📚 Technical research
🔄 Migration planning
⚙️ System design
🤖 AI agent workflows

These are problems where a second or third perspective can expose something the first model overlooked.

⚖️ The Trade-Offs

Mixture of Agents isn’t a free intelligence upgrade.

💰 Higher API Costs

You’re making additional model calls.

With two reference models and one aggregator, you’re paying for multiple perspectives instead of relying entirely on one model.

Hermes preserves prompt caching, which helps, but the additional reference calls still introduce extra cost.

⏳ Increased Latency

The aggregator needs the reference outputs before it can use them.

That means complex tasks can take longer than simply sending everything directly to one model.

🤔 More Models ≠ Better Results

Adding models blindly isn’t the goal.

If all your reference models have similar strengths and approach problems similarly, the additional perspectives may not add much.

The real value comes from diversity.

A strong combination might include models that excel at different things rather than simply choosing the three highest-scoring models on a leaderboard.

💡 The Bigger Idea: AI Models Don’t Have to Compete

The most interesting takeaway for me wasn’t the Kanban application itself.

It was the change in how we can think about AI models.

Most of the AI industry conversation revolves around competition:

Which model is #1?

Which model has the highest benchmark score?

Which model should replace the one I’m currently using?

But maybe that’s only part of the story.

The next step in AI systems might not simply be:

One smarter model.

It could also be:

Multiple specialized models working together.

Instead of choosing between Claude, DeepSeek, MiniMax, GPT, or whatever comes next, AI agents can potentially use different models as a team—taking advantage of their individual strengths before making a final decision.

The question then changes from:

“Which AI model is the smartest?”

to:

“Which combination of AI models makes the smartest decisions?”

And I think that’s a much more interesting problem.

💬 Which 3 AI Models Would You Choose?

Here’s the experiment I’d love to see:

You get to build your own AI team.

You can choose two advisors and one final decision-maker.

Which three models are you picking?

And more importantly:

Why that combination?

Drop your AI team in the comments. 👇

🚀 I Built an AI Company That Shipped an App From Scratch (Using Paperclip AI)

Vivek Shetye — Tue, 07 Jul 2026 16:56:07 +0000

💬 What if AI didn’t work as a single assistant but as an entire software company?

That’s exactly what I wanted to explore.

Instead of asking one AI agent to build an application, I created an AI engineering organization with specialized roles and let them collaborate like a real development team.

The result?

A working CRM Dashboard MVP built using Paperclip AI, complete with planning, architecture, implementation, testing, security reviews, and multiple UI refinement iterations.

In this article, I’ll show you how the workflow works, what impressed me, where it still falls short, and why I think this is one of the most interesting directions for AI-powered software development.

🎥 Full video walkthrough

🏢 Building an AI Company

Most AI coding tools give you a single powerful agent.

Paperclip AI takes a different approach.

Instead of one “super agent,” it lets you build an organization with reporting structures, responsibilities, approvals, and workflows that resemble a real engineering team.

For this project, my company consisted of:

👔 CEO
🏗️ CTO
💻 Software Engineer
🧪 QA Engineer
🛡️ Security Engineer

Every agent had:

Clearly defined responsibilities
Its own Agents.md
Its own SOUL.md
Specialized models and tools
A reporting hierarchy

Rather than sharing one giant prompt, every agent knew exactly what it was responsible for.

🎯 The Project

To test the workflow, I asked my AI company to build a production-ready CRM Dashboard MVP.

The application includes:

🔐 Secure authentication
📂 CSV lead import
📊 Visual sales pipeline
🔍 Search & filtering
📈 Dashboard metrics
📝 Opportunity management
📱 Responsive UI

Instead of writing code myself, I focused on defining requirements and improving the workflow.

⚙️ How the Workflow Actually Worked

This was probably the most interesting part.

The CEO didn’t immediately start generating code.

Instead, it:

Analyzed requirements
Broke down the project
Created an execution plan
Defined dependencies
Requested approval

Only after the roadmap was approved did the CTO begin working.

The CTO:

Designed the system architecture
Selected the technology stack
Planned APIs
Designed the database
Defined security requirements

Only then did implementation begin.

The Software Engineer implemented features.

After each implementation:

🧪 QA validated functionality.
🛡️ Security reviewed vulnerabilities.
🏗️ The CTO reviewed and approved completed work before tasks were marked finished.

It felt much closer to how real engineering organizations operate than traditional “single prompt” AI coding.

🎨 Iterating Instead of Starting Over

One thing I enjoyed was how easy it was to improve the application incrementally.

The initial dashboard worked…

…but the UI looked cramped and inconsistent.

Instead of regenerating everything, I created another task with:

acceptance criteria
design improvements
spacing fixes
layout enhancements
interaction improvements

The AI engineering team iterated on the existing application just like a real software team would.

🤖 Different Models for Different Jobs

Another thing I liked was assigning different models to different roles.

For example:

👔 CEO → Hermes
🏗️ CTO → Hermes
💻 Software Engineer → Codex (GPT-5.5)
🧪 QA → Codex (GPT-5.5)
🛡️ Security → Codex (GPT-5.4)

Instead of expecting one model to excel at everything, each agent could specialize.

That approach feels much more scalable as AI models continue improving.

💡 What Impressed Me Most

It wasn’t that AI generated code.

We’ve already seen that.

What impressed me was how work flowed through the organization.

Requirements.

↓

Planning.

↓

Architecture.

↓

Implementation.

↓

Testing.

↓

Security review.

↓

Approval.

↓

Iteration.

That mirrors the software development lifecycle surprisingly well.

⚠️ Let’s Set Realistic Expectations

Paperclip AI isn’t magic.

It won’t build a production SaaS application from a single prompt.

The quality depends heavily on:

Choosing the right model for each role
Writing good Agents.md
Designing strong SOUL.md
Creating clear task dependencies
Installing useful skills
Reviewing outputs
Iterating frequently

Think of it as managing an engineering team rather than using an autocomplete tool.

The better your team is organized, the better the results become.

🚀 Final Thoughts

I don’t think the future of AI software engineering is one massive agent doing everything.

I think it’s teams of specialized AI agents collaborating through structured workflows, with humans acting as engineering managers rather than code generators.

Paperclip AI is still evolving, and there are definitely rough edges.

But after building this project, I’m convinced that structured multi-agent orchestration is a direction worth watching.

I’m excited to see where it goes next.

If you’ve experimented with Paperclip AI, AI agents, or multi-agent software engineering, I’d love to hear your experience in the comments.

Happy building! 🚀

🤖 I Built an AI Company With Paperclip AI

Vivek Shetye — Mon, 29 Jun 2026 19:51:33 +0000

What if your next startup didn’t begin with hiring developers…

What if it began with hiring AI employees?

Not just one AI assistant.

An actual company.

→ A CEO that plans the roadmap.
→ A CTO that designs the architecture.
→ Software Engineers that write production code.
→ QA Engineers that test every feature.
→ Security Engineers that review vulnerabilities.
→ A Marketing Manager that prepares the launch.

And instead of you manually coordinating everything, they collaborate, delegate work, request approvals, and execute projects together.

That sounds like science fiction…

Until I discovered Paperclip AI.

🚀 Why Paperclip AI Feels Different

Most AI frameworks are trying to answer one question:

“How can we make one AI agent smarter?”

Paperclip asks a completely different question:

“How can multiple AI agents work together like a real company?”

That simple shift completely changes how you think about building AI systems.

Instead of creating a super-agent that does everything, you build specialized employees that each focus on what they’re best at.

It’s organizational intelligence instead of individual intelligence.

🎥 Full video walkthrough

🏢 Think of Paperclip as Your AI Company

One analogy made everything click for me.

Imagine these popular coding agents are individual employees:

Claude Code
Codex
Hermes Agent
OpenClaw

Paperclip isn’t competing with them.

Paperclip manages them.

It becomes the operating system for your AI workforce.

Each employee can even use a different backend depending on what they’re good at.

For example:

👨‍💼 CEO → Hermes Agent

🛡️ Security Engineer → Codex

🎨 Designer → Claude Code

You don’t have to force one agent harness to do everything anymore.

You simply choose the best employee for each job.

⚡ The AI Company I Built

For this demo I built an organization with multiple AI employees.

CEO
├── CTO
│   ├── Full Stack Engineer
│   ├── QA Engineer
│   ├── Security Engineer
│   └── DevOps Engineer
└── CMO

Every employee has:

📋 Their own dashboard
🧠 Individual instructions
🛠️ Custom skills
📊 Activity history
💰 Budget limits
🔒 Permission controls
🤝 Ability to assign work

Watching these agents coordinate with each other feels surprisingly similar to managing a real engineering team.

🎯 Features That Really Impressed Me

🧩 Mix Different Agent Harnesses

Every employee can run on a completely different backend.

You’re no longer locked into one provider.

Use Claude where it’s strongest.

Use Codex where it’s strongest.

Use Hermes Agent where it excels.

Mix and match however you like.

📂 Projects

Instead of hundreds of unrelated tasks, everything belongs to a project.

Projects contain:

Goals
Tasks
Deliverables
Assigned agents

Which makes tracking large software projects much easier.

🎯 Goals

Goals define long-term company objectives.

Instead of random prompts, your entire AI organization stays aligned toward a shared mission.

🔄 Routines

One of my favorite features.

You can schedule recurring work such as:

Daily research
Weekly code reviews
Security scans
Competitor monitoring
Lead generation

Your AI employees wake up automatically and continue working without manual intervention.

📦 Artifacts

Every report…

Every design…

Every document…

Every generated file…

Gets stored in one place so your AI company builds a growing knowledge base over time.

💸 Budget Controls

Probably one of the most practical features.

You can define spending limits for every employee.

If an agent reaches its budget…

It stops.

No surprise API bills.

No infinite execution loops.

Just predictable automation.

⚙️ Installing Paperclip AI

Getting started was surprisingly simple.

The installation is essentially:

Install Paperclip
Give your company a name
Define your mission
Create your CEO
Choose an AI backend
Assign the first task

Within minutes…

You already have your first AI company running.

💡 What Changed My Perspective

Before Paperclip, I thought about AI as individual assistants.

After using it…

I started thinking about AI organizations.

The future probably isn’t one super-intelligent agent doing everything.

It’s multiple specialized agents collaborating like real teams.

That feels much closer to how successful companies actually work.

🚀 What’s Next?

This is only the beginning.

In the next videos I’ll be using Paperclip AI to build a real software startup from scratch.

We’ll have AI employees planning products, writing code, testing features, fixing bugs, and even generating go-to-market strategies.

I’m excited to see just how far this concept can go.

💬 What Would Your AI Company Look Like?

If you could hire only one AI employee today…

Who would you choose first?

👨‍💼 CEO

🏗️ CTO

💻 Software Engineer

🧪 QA Engineer

📈 Marketing Manager

Or something completely different?

I’d love to hear your ideas.

🚀 I Built an AI Team That Finds Leads, Decision Makers & Writes Outreach While I Sleep

Vivek Shetye — Tue, 23 Jun 2026 19:26:26 +0000

Lead generation is one of those tasks everyone knows they should automate… but most teams still do it manually.

Every week the process looks something like this:

🔎 Search LinkedIn for companies

🌐 Visit dozens of websites

👤 Find the CTO, VP, or Director

📧 Hunt for contact information

✍️ Write personalized outreach emails

⏳ Lose half a day doing repetitive work

What if all of that happened automatically?

What if you woke up Monday morning to find:

✅ A list of companies matching your ICP

✅ Decision makers already identified

✅ Company research already completed

✅ Lead qualification scores calculated

✅ Personalized outreach drafts ready to send

That’s exactly what I built using Hermes Agent.

Full Video Walkthrough:

🤖 The Goal

I wanted a system where I could provide a simple campaign brief and have a team of AI agents handle the entire lead generation workflow.

Something like:

“Find SaaS startups with 10-200 employees that build AI developer tools. Identify decision makers and prepare personalized outreach.”

Instead of manually managing every step, a multi-agent workflow handles the process from start to finish.

🏗️ Architecture Overview

The pipeline consists of six specialized AI agents:

🎯 Orchestrator Agent

Acts like a sales manager.

Responsibilities:

Creates execution plans
Creates tasks
Assigns work to specialist agents
Tracks progress
Manages dependencies
Controls workflow phases

🔍 Prospector Agent

Finds companies that match your ICP.

Input:

Keywords
Industry
Company size
Target market

Output:

Qualified company list

Example keywords:

AI Developer Tools
LLM Infrastructure
DevOps Automation
AI Engineering Platforms

🌐 Scraper Agent

Researches every company discovered during prospecting.

Collects:

Products
Services
Company descriptions
Locations
Social profiles
Technology signals

All data gets enriched and stored automatically.

👥 Contact Finder Agent

Identifies the right people inside each company.

Targets:

CTOs
VPs
Directors
Founders

Then gathers available contact information from multiple sources.

✉️ Outreach Agent

Generates personalized outreach emails.

Instead of generic templates, it references:

Company initiatives
Product offerings
Technology stack
Industry positioning

Result:

Much more relevant outreach messages.

📊 Analyst Agent

Scores every lead against the Ideal Customer Profile (ICP).

Evaluation criteria:

Company size
Industry fit
Product relevance
Buying potential
Strategic alignment

Each company receives a qualification score between 0 and 1.

This helps prioritize outreach efforts.

🧠 Why Multi-Agent Systems Work So Well

Most people try to build lead generation using a single AI agent.

The problem?

One agent becomes responsible for:

Research
Scraping
Qualification
Personalization
Coordination

That quickly becomes messy.

Instead, I use specialist agents.

Each agent focuses on one responsibility only.

Benefits:

✅ Better task quality

✅ Easier debugging

✅ Better scalability

✅ Parallel execution

✅ Cleaner workflows

📋 Workflow State Machine

The workflow runs in phases:

Campaign Brief
      │
      ▼
 Prospecting
      │
      ▼
 Research & Enrichment
      │
      ▼
 Contact Discovery
      │
      ▼
 Outreach Generation
      │
      ▼
 Lead Qualification
      │
      ▼
 Campaign Report

The orchestrator only unlocks the next phase once the previous phase is completed successfully.

This prevents bad downstream data from contaminating later stages.

🗂️ Hermes KANBAN Board = Shared Agent Memory

One of my favorite parts of Hermes Agent is its Kanban workflow system.

The Kanban board acts as a shared coordination layer between agents.

Every agent can:

📖 Read task status

✍️ Update progress

🔄 Create follow-up tasks

🚦 Track dependencies

The orchestrator uses the board to understand:

What is complete
What is blocked
What should happen next

This creates a surprisingly robust autonomous workflow.

⚡ Running a Campaign

To launch a campaign I simply provide:

Product

What I’m selling

ICP

Who I want to target

Discovery Keywords

Where prospecting should begin

Goal

What outcome I want

Example:

Run a full B2B lead generation campaign for ShipMe Agent, an AI agent for QA and DevOps automation. 
ICP: AI SaaS startups, 10-200 employees. 
Keywords: AI Developer Tools, LLM DevOps Automation. 
Goal: 10 ranked qualified leads with decision-maker contacts and personalized outreach emails drafted

The orchestrator takes over from there.

📈 End Result

At the end of a run I receive all information organized in CSV files:

Companies

ICP matched companies
Enriched company data

Contacts

Decision makers
Contact information

Outreach

Personalized email drafts

Qualification

Lead scoring
Prioritized opportunities

Report

Campaign summary
Workflow results

All generated automatically from a single task.

🔥 Real Value

The biggest win isn’t saving a few minutes.

It’s eliminating repetitive work entirely.

Instead of spending hours every week:

❌ Searching

❌ Researching

❌ Copy-pasting

❌ Writing first drafts

I can focus on:

✅ Sales conversations

✅ Closing deals

✅ Improving campaigns

✅ Building relationships

The AI team handles the operational work.

🛠️ Resources

🔗 Agents SOUL.md files & Custom Hermes Skills: https://github.com/vivekshetye/hermes-lead-generation-pipeline

💬 What Would You Automate?

If you had a team of AI agents working for you 24/7, what business workflow would you automate first?

Lead generation?

Customer support?

Market research?

Content creation?

I’d love to hear what you’re building.

🚀 Hermes Agent Just Released a Desktop App And It Changes Everything About Using AI Agents

Vivek Shetye — Thu, 18 Jun 2026 20:41:34 +0000

For the last year, AI agents have been getting more powerful.

The problem?

Most of them still felt like developer tools.

You had to work in terminals, manage configuration files, memorize commands, and scroll through endless logs just to understand what your agent was doing.

That’s fine if you’re a developer.

But if AI agents are going to become mainstream, they need something else:

A great user experience.

That’s exactly what Hermes Agent’s newly released Desktop App delivers.

🎥 Full video walkthrough

🤔 The Biggest Problem With AI Agents

Most AI agent frameworks have incredible capabilities:

Browse the web
Read and write files
Execute code
Automate workflows
Use multiple tools
Coordinate sub-agents

But using them often looks like this:

❌ Terminal windows everywhere
❌ Session IDs to manage manually
❌ Configuration files to edit
❌ Logs that are difficult to understand
❌ Very little visibility into what the agent is actually doing

For experienced developers, this is manageable.

For everyone else, it’s a major barrier.

🧠 What Is Hermes Agent?

Hermes Agent is an open-source AI agent framework created by the team at Nous Research.

It can run locally on your machine or on remote servers and connect to multiple AI providers including:

🧩 OpenAI
🌐 Gemini
🧠 Claude
🏠 Local models through Ollama
🔌 Other supported providers

Once connected, Hermes can:

Browse the web
Analyze documents
Run terminal commands
Automate workflows
Send messages
Manage emails
Create and execute multi-step plans

Think of it as an autonomous AI worker that can actually take action instead of only generating text.

✨ The Desktop App Changes Everything

The new desktop application provides a complete visual interface for managing and monitoring AI agents.

Instead of wondering what your agent is doing, you can now see everything.

📂 Session Management

Conversations are automatically organized.

Sessions are grouped by profile, making it much easier to manage multiple agents with different responsibilities.

You can also switch models with a single click without diving into configuration files.

🔍 Watch Your Agent Work

One of the most impressive features is transparency.

You can inspect:

🔎 Tool calls
📚 Sources used by the agent
⚙️ Workflow execution steps
🧠 Reasoning process
📈 Agent progress

This is incredibly useful when debugging workflows or understanding why an agent made a particular decision.

Most AI agent platforms treat this as a black box.

Hermes makes it visible.

🎙️ Voice Interaction

Hermes supports voice input directly from the desktop app.

You can simply speak to your agent and let the local transcription system convert speech into text.

Small feature.

Huge usability improvement.

👤 Profiles: Specialized AI Agents

This might be my favorite feature.

A Hermes Profile is essentially its own AI agent.

Each profile can have:

📝 Independent instructions
🧠 Separate memory
🛠️ Different tools
📚 Different skills
⚡ Unique capabilities

For example:

Software Engineering Agent
Research Agent
Content Creation Agent
Marketing Agent
Stock Research Agent

Instead of one general-purpose assistant, you can build an entire team of specialized AI workers.

🛠️ Skills and Tools

Hermes includes a powerful skill system.

What’s particularly interesting is that Hermes can generate new skills from your conversations over time.

The more you use it, the more personalized it becomes.

You can also selectively disable skills to:

🎯 Reduce context size
💰 Save tokens
⚡ Improve performance

This is a subtle feature that becomes very important at scale.

💬 Messaging Integrations

The desktop app supports integrations with external platforms such as:

💬 Discord
📱 Telegram
📨 WhatsApp
🔗 Other supported channels

Your agents can communicate and deliver updates outside the desktop application itself.

This opens the door to some very interesting automation workflows.

📦 Artifacts: Everything In One Place

One challenge with AI agents is finding things they created days ago.

Hermes solves this with Artifacts.

Generated files, images, links, and outputs are automatically collected into a centralized workspace.

No more hunting through old conversations.

⚙️ Advanced Settings For Power Users

The settings panel exposes a surprising amount of customization.

You can configure:

⚙️ AI Providers
🔑 API Keys
🎨 Appearance
🔌 MCP Integrations
🎙️ Voice Settings
🛠️ Tool Configuration
🌐 Gateway Settings

You can even assign different models to different tasks.

For example:

One model for reasoning
Another for vision
Another for web extraction

This level of flexibility is something advanced users will appreciate.

⏰ Autonomous Scheduled Agents

One feature I think is massively underrated:

📅 Cron Jobs

You can schedule agents to run automatically.

Examples:

📈 Daily stock market reports
📧 Email summaries
📰 Industry news monitoring
🏢 Competitor tracking
📊 Business intelligence reports

You define:

The task
The schedule
The delivery destination

Then Hermes runs it automatically.

Your AI agent becomes proactive instead of reactive.

👥 Multi-Agent Visibility

When Hermes encounters a complex task, it can spawn additional agents to help complete the work.

The Desktop App includes a dedicated view for monitoring these sub-agents.

You can watch:

👥 Which agents were created
📋 What tasks they are handling
🔄 Their current progress
🎯 How work is being coordinated

For anyone interested in multi-agent systems, this is fascinating to observe.

🎯 Why This Matters

The most important thing about this release isn’t the interface itself.

It’s what it represents.

AI agents are moving from:

“Developer-only tools”

“Tools anyone can use.”

The Desktop App dramatically lowers the barrier to entry while preserving the power that makes Hermes compelling.

And that’s exactly what the AI agent ecosystem needs right now.

What feature do you think is still missing from modern AI agent platforms? Let me know in the comments.

🚀 Build a Fully Local AI Agent with Hermes Agent, Ollama, Qwen 3.5, and SearXNG (100% Private & $0 Cost)

Vivek Shetye — Mon, 08 Jun 2026 20:02:58 +0000

What if you could build an AI agent that can:

✅ Think and reason

✅ Search the web

✅ Read and write files

✅ Generate reports and dashboards

✅ Run entirely on your own machine

Without:

❌ OpenAI API keys

❌ Anthropic subscriptions

❌ Monthly AI bills

❌ Sending your prompts and files to third-party servers

That's exactly what I built.

In this tutorial, I'll show you how to create a fully local AI agent stack using:

🤖 Hermes Agent

🧠 Qwen 3.5 9B via Ollama

🔎 SearXNG

The result is a powerful AI agent that costs $0 to operate, keeps your data private, and gives you complete control over your AI infrastructure.

🎥Full video walkthrough:

🤔 Why Build a Local AI Agent?

Most AI agents today depend on cloud APIs.

Every prompt, file, and conversation gets sent to someone else's servers.

For many use cases, that's perfectly fine.

But what if you're working with:

🔒 Sensitive business information

🔒 Private research data

🔒 Customer documents

🔒 Internal company knowledge

🔒 Personal notes and files

In those scenarios, privacy matters.

A local AI agent means:

✅ Your data never leaves your machine

✅ No third-party access to your prompts

✅ No API costs

✅ No rate limits

✅ Full ownership of your stack

And thanks to modern open-source models, local AI is becoming surprisingly capable.

🏗️ The Architecture

Our stack consists of three components.

🤖 Hermes Agent

Hermes Agent is an open-source AI agent framework developed by Nous Research.

Instead of just chatting with an LLM, Hermes turns the model into a true agent with:

Memory
Tool usage
Workflows
File access
Web search
Task execution

Think of it as the operating system for your AI agent.

🧠 Qwen 3.5 9B via Ollama

Next comes the brain.

We're using Qwen 3.5 9B running locally through Ollama.

Ollama makes it incredibly easy to run modern open-source language models on your machine.

The model handles:

Reasoning
Planning
Decision making
Report generation
Tool calling

And because it's running locally, every token stays on your hardware.

🔎 SearXNG

The final piece is SearXNG.

SearXNG is a privacy-focused meta search engine.

Instead of tracking users like traditional search providers, it aggregates results from multiple search sources while preserving privacy.

For AI agents, this means:

✅ Web search capabilities

✅ No tracking

✅ Self-hosted infrastructure

✅ Complete control

⚡ What Makes This Stack Interesting?

Most developers assume AI agents require expensive cloud infrastructure.

But with this setup:

💰 API Cost = $0

🔒 Data Privacy = 100%

⚙️ Infrastructure Ownership = 100%

🛠️ Customization = Unlimited

Everything runs locally.

Everything remains under your control.

🎯 Real Demo

To test the setup, I gave the agent a simple task:

Find the latest AI news and create an HTML report.

Here's what happened.

Step 1

The agent used SearXNG to search the web.

Step 2

It gathered and synthesized information from multiple sources.

Step 3

It generated a structured HTML report.

Step 4

The file was saved locally on my machine.

No cloud APIs.

No external AI providers.

No third-party processing.

Just a fully local AI agent doing real work.

🔥 The Best Part: It Scales

One thing I love about this architecture is that it grows with your hardware.

Starting point:

🧠 Qwen 3.5 9B

Future upgrades:

🚀 Larger Qwen models

🚀 70B parameter models

🚀 400B parameter models

🚀 Multi-GPU setups

The architecture stays exactly the same.

You simply swap in a more capable model.

The only real limitation is your hardware.

💡 Potential Use Cases

Developers are already building some fascinating things with local AI agents.

Examples include:

📚 Research assistants

📄 Private document analysis

💻 Coding assistants

📈 Market research workflows

📰 News aggregation systems

📋 Report generation pipelines

🏢 Internal company knowledge assistants

🔬 Scientific research agents

🔒 Privacy-first enterprise AI solutions

Because everything is self-hosted, these use cases become much easier to justify from a security and compliance perspective.

🌍 Why Local AI Is Becoming a Big Deal

The AI industry spent the last few years moving everything to the cloud.

Now we're seeing another trend emerge:

Bringing AI back to the device.

Open-source models are improving rapidly.

Consumer hardware is becoming more powerful.

Agent frameworks are becoming more capable.

As a result, local AI is no longer just a hobby project.

It's becoming a practical option for real-world applications.

The combination of:

🤖 AI Agents

🧠 Open Models

🔒 Privacy

💰 Zero API Cost

is incredibly compelling.

💬 What Would You Build?

If you had a fully private AI agent running entirely on your own machine...

What would you build?

A coding assistant?

A research agent?

A private knowledge system?

A business automation workflow?

Let me know in the comments. I'm always curious to see what developers are creating with local AI.

🚀 Hermes Agent Just Added a Kanban Board — And It Changes Everything

Vivek Shetye — Thu, 04 Jun 2026 18:59:26 +0000

Most AI agent frameworks have a problem.

You can create agents.

You can give them tools.

You can even make them collaborate.

But as soon as your workflow gets complex, things start breaking down.

📌 Tasks become difficult to track.

📌 Dependencies become messy.

📌 You lose visibility into what’s running.

📌 Outputs become disconnected.

📌 Coordination becomes manual.

The new Hermes Agent Kanban Board solves exactly that problem.

And after spending time with it, I genuinely think this is one of the biggest upgrades Hermes Agent has shipped so far.

To test it, I decided to orchestrate an entire product launch campaign using multiple specialized AI agents working together inside a single workflow.

What happened was pretty impressive.

🤔 Why Most AI Workflows Fall Apart

Most AI tools today are conversation-driven.

The workflow usually looks like this:

💬 Open a chat

✍️ Write a prompt

📄 Copy the output

📋 Paste it somewhere else

🔄 Repeat 20 more times

Eventually you’re juggling:

Multiple chat windows
Notion pages
Google Docs
Spreadsheets
Task trackers

The AI may be smart.

But the workflow isn’t.

And that’s where Hermes Kanban enters the picture.

🎯 What Exactly Is Hermes Kanban?

Hermes Kanban is a visual workflow orchestration system built directly into Hermes Agent.

Instead of managing agents through separate chats, you coordinate everything through a task board.

You get:

✅ Task assignment

✅ Dependencies

✅ Parent-child relationships

✅ Parallel execution

✅ Live logs

✅ Workflow visualization

✅ Artifact management

Think of it as:

🧠 AI Agents + 📋 Trello + 🎯 Workflow Automation

all inside a single system.

🎥 Watch The Full Walkthrough

In the video, I cover:

✅ Installing Hermes Agent

✅ Creating specialist agents

✅ Setting up web search

✅ Building Kanban workflows

✅ Managing dependencies

✅ Running tasks in parallel

✅ Reviewing outputs

✅ Publishing final deliverables

🔥 The Feature That Immediately Got My Attention

Task Dependencies.

This sounds simple.

But it’s incredibly powerful.

Imagine you’re launching a product.

Before creating content, you need:

🔍 Audience Research

Then:

📊 Messaging & Positioning

Only after that can you create:

✍️ Blog Posts

📧 Email Sequences

📱 Social Content

🎥 YouTube Scripts

With Hermes Kanban, those relationships are explicitly defined.

Audience Research
       ↓
Messaging & Positioning
       ↓

Landing Page
Blog Posts
Emails
Social Posts
YouTube Scripts

Every downstream task can reference outputs generated by upstream tasks.

This creates a true workflow rather than a collection of isolated prompts.

⚡ Parallel Agent Execution Is Where Things Get Interesting

Once positioning was complete, I launched five content-generation tasks simultaneously.

Suddenly I had:

✍️ Landing Page Creation

📝 Blog Writing

📧 Email Sequence Generation

📱 Social Media Content

🎥 YouTube Script Creation

all running at the same time.

Not one after another.

Not manually triggered.

The system orchestrated everything automatically.

Watching multiple AI agents execute work in parallel felt less like using an AI tool and more like managing a real team.

👥 Building a Team of Specialized AI Agents

For this demo, I created four specialist agents.

🔬 Researcher

Responsible for:

Competitor analysis
Audience discovery
Market research
Pain point analysis

⸻

📊 Analyst

Responsible for:

Positioning
Messaging
Strategic narratives
Market differentiation

⸻

✍️ Writer

Responsible for:

Landing pages
Blogs
Emails
Social content
Video scripts

⸻

🔍 Reviewer

Responsible for:

Quality control
Consistency checks
Accuracy validation
Brand alignment

Each agent had its own:

🧠 Memory

⚙️ Configuration

📖 Instructions

🎯 Responsibilities

This specialization dramatically improves output quality.

👀 Seeing What Your Agents Are Doing

One thing I really appreciated was transparency.

Every task provides:

📜 Execution logs

📊 Status updates

📂 Generated artifacts

⏱️ Progress tracking

Instead of wondering:

“What is my agent doing right now?”

You can actually see it.

For complex workflows, this is incredibly valuable.

🚀 My Demo Workflow: Launching Momentum

To showcase the Kanban board, I created a launch workflow for a fictional product called Momentum.

Momentum is an AI-powered habit tracker that adapts to your natural energy patterns.

The workflow looked like this:

🔍 Audience Research

↓

📊 Messaging & Positioning

↓

✍️ Landing Page

📝 Blog Posts

📧 Email Sequence

📱 Social Posts

🎥 YouTube Scripts

↓

🔍 Content Review

↓

✅ Final Revisions

The marketing campaign itself wasn’t the interesting part.

The interesting part was watching Hermes Kanban coordinate the entire process.

🧩 The Reviewer Agent Was Surprisingly Useful

Once all content was generated, I handed everything over to the Reviewer agent.

It analyzed:

📧 Emails

📄 Landing Pages

📝 Blog Content

📱 Social Posts

🎥 Scripts

and checked for:

✅ Consistency

✅ Clarity

✅ Accuracy

✅ Tone

It identified several issues and produced actionable feedback.

Then I created one final task.

The Writer agent consumed that feedback and automatically updated all assets.

This created a genuine feedback loop between agents.

Exactly how real teams operate.

🌎 This Goes Far Beyond Marketing

The marketing workflow was simply a demonstration.

The same approach could be used for:

💻 Software Development

Research → Design → Code → Review → Documentation

📈 SEO Operations

Keyword Research → Content Brief → Writing → Optimization

🧪 Research Projects

Data Collection → Analysis → Reporting

🚀 Startup Execution

Research → Strategy → Outreach → Growth

📰 Content Teams

Research → Writing → Editing → Publishing

Anywhere you have a repeatable workflow, Hermes Kanban becomes interesting.

💡 Why I Think Hermes Kanban Is A Big Deal

Most AI products focus on conversations.

Hermes Kanban focuses on workflows.

That’s a major difference.

Instead of:

👤 Human → 🤖 AI

you get:

👤 Human → 👥 AI Team

One agent researches.

One analyzes.

One writes.

One reviews.

The workflow coordinates everything.

You supervise.

That feels much closer to the future of AI systems than simply chatting with a chatbot.

💬 What Would You Build?

The most exciting part of Hermes Kanban isn’t marketing automation.

It’s the ability to coordinate teams of AI agents through structured workflows.

I’m curious:

👉 What workflow would you automate first?

Product development?

Content operations?

Research?

Customer support?

Let me know in the comments.

🚀 Build a Self-Improving AI Assistant with Hermes (Beginner-Friendly Step-by-Step Guide)

Vivek Shetye — Thu, 07 May 2026 19:26:38 +0000

Most AI tools today are still glorified chatbots.

You ask a question.
You get an answer.
And by tomorrow… it forgets you even exist.

But what if you could build an AI assistant that:

🧠 Remembers your preferences
📈 Learns from every interaction
🌐 Researches the web in real time
📱 Runs directly on your phone
⏰ Automates tasks while you sleep

That’s exactly what I built using Hermes, one of the fastest-rising AI agent frameworks right now and increasingly seen as a serious competitor in the personal AI space.

🔥 Why Hermes Is Blowing Up

Hermes isn’t just another chatbot framework.

It introduces something much more powerful:

🧠 Long-Term Memory

Your assistant remembers:

Your goals
Your recurring tasks
Your preferences
Your habits

🔄 Self-Reflection

After every task, Hermes can evaluate:

What worked
What failed
What should improve next time

This means your AI literally gets smarter the more you use it.

🤖 Real Automation

You can use Hermes for:

Daily AI news briefings
Travel planning
Productivity reminders
Research tasks
Personal scheduling
Coding assistant
Messaging integrations like Telegram,Discord, etc

🛠️ What You’ll Build in This Tutorial

In this beginner-friendly walkthrough, I cover:

✅ Installing Hermes from scratch
✅ Configuring free Google AI Studio API access
✅ Adding Tavily for live web search
✅ Connecting your AI assistant to Telegram
✅ Creating a true phone-based personal assistant
✅ Security best practices

🎥 Full video walkthrough

⚙️ Installation: Build Your First Hermes AI Assistant

Getting started with Hermes is surprisingly simple — even if you’re completely new to AI agents.

Within minutes, you’ll have your own self-improving assistant running locally.

🚀 Step 1: Run the Hermes Installer

Simply run:

curl -fsSL https://raw.githubusercontent.com/NousResearch/hermes-agent/main/scripts/install.sh | bash

What this does:

✅ Downloads Hermes
✅ Installs required dependencies
✅ Configures your environment
✅ Guides you through setup interactively

💡Pro Tip: macOS and Linux users can install directly. Windows users should first install WSL2 for the smoothest experience.

🤖 Step 2: Choose Your Model Provider

Hermes supports multiple providers, but for beginners, Google AI Studio (Gemini) is one of the best choices because:

💸 Generous free tier
⚡ Fast setup
🛠️ Beginner-friendly
🚀 Great for experimentation

During setup, select:

(•) Google AI Studio (Gemini models — native Gemini API)

🔑 Step 3: Get Your Google AI Studio API Key

To power Hermes, you’ll need an API key.

Steps:

Go to Google AI Studio
Navigate to Projects
Click Create New Project
Go to API Keys
Select your project
Click Create API Key
Copy your key
Paste it into the Hermes installer when prompted

🔒 Security Tip:
Treat API keys like passwords — never share them publicly.

📱 Step 4: Skip Messaging Setup (For Now)

During initial installation, Hermes will ask about messaging integrations.

👉 Skip this step temporarily.

We’ll configure Telegram later to transform Hermes into a true mobile personal assistant.

💬 Step 5: Launch Hermes Chat

When prompted, enter: y

Hermes will launch its built-in TUI (Terminal User Interface).

You now have:

✅ Your first AI agent
✅ Local chat interface
✅ Long-term memory capabilities
✅ Self-improvement foundation

🎉 Congratulations — your personal AI assistant is officially alive.

🔍 Configure Tavily for Real-Time Web Search

Without web access, your AI assistant is limited.

Adding Tavily gives Hermes the ability to:

🌐 Search the internet
📰 Gather live information
📊 Perform research tasks
✈️ Plan trips using current data

🛠️ Tavily Setup Steps

Create a Tavily account

Visit Tavily and sign up.

Generate your API key

Create a new key from your dashboard.

Add it to Hermes

hermes config set TAVILY_API_KEY <YOUR_TAVILY_API_KEY>

🎯 Result:

Your Hermes assistant can now perform real-time web searches and research autonomously.

📲 Connect Hermes to Telegram (Your AI in Your Pocket)

Right now, Hermes only runs on your computer.

Connecting Telegram transforms it into a true personal assistant you can message from anywhere.

🤖 Step 1: Create a Telegram Bot

Instructions:

Open Telegram
Search for @BotFather
Type: /newbot
Choose a bot name
Choose a unique username ending in: _bot
Copy your bot token

🔑 You’ll need this token to connect Hermes.

Step 2: Connect Telegram to Hermes

Run the following command:

hermes setup gateway

Then:

✅ Select Telegram
✅ Paste your Bot Token
✅ Add your Telegram numeric user ID
✅ Set your account as the home channel
✅ Start the gateway service

📍 How to Find Your Telegram User ID

Search Telegram for: @userinfobot

It will instantly provide your numeric user ID.

🎉 Final Result

Your Hermes AI Assistant is now:

📱 Available on your phone
💬 Reachable through Telegram
🧠 Memory-enabled
🌐 Web-connected
📈 Self-improving

You’ve officially built a beginner-friendly AI agent that can evolve into a true personal productivity system.

🌍 Real-World Examples

Once set up, your AI assistant can:

✈️ Personalized Travel Planning

Ask it to plan a trip to Japan… then later France…

And it remembers:

Your budget
Your interests
Your travel style

📰 Daily Briefings

Example:

“Every morning at 8 AM, send me the top 3 AI news stories.”

Your assistant works while you sleep.

📅 Weekly Productivity Systems

Example:

“Every Friday at 4 PM, remind me to review my week.”

⚠️ Important Security Tips

Hermes is powerful — but with power comes responsibility.

Best Practices:

🔒 Avoid running on your primary machine
🖥️ Use a VM, remote server, or sandbox
🔑 Protect API keys like passwords
⚡ Be cautious with community-built skills

🧠 Why This Is Bigger Than Just One Tutorial

We’re watching AI evolve from:

Old Model:

Chatbot → Answer → Forget

New Model:

Assistant → Learn → Remember → Improve

This is a massive shift.

We’re entering the era of persistent personal AI.

💬 Final Thoughts

Hermes shows where AI is truly heading:

✨ Personalized
✨ Persistent
✨ Autonomous
✨ Self-improving

For beginners, this may be one of the most practical AI agent projects you can build right now.