I created a music production tool and am working with artists to create marketing videos for it. I get back a bunch of loosely related video files and need to put together a coherent video for marketing and ads.

For this I need to:

• Come up with a cohesive story for the video

• Edit the video

• Make variants so that we can test out different versions for paid advertising

This can be pretty time consuming and it’s hard to keep track of what variants are possible if you get too deep into the footage. It’s better to work at a higher level to define the variants and then get into the details on each video.

So I made an AI agent that does this. I point it at my raw footage archive, give it some loose directions and then it can come up with some video concepts, mock them up, and let me decide which directions I want to pursue and refine into videos.

My Agent

How it works

• Simple version: the agent runs on my computer

  • (Tech nerd version: I run OpenClaw in an Ubuntu VM on my NAS and I have a video editor sub-agent)

• It works on files saved on a folder synced to the cloud and this whole folder syncs across to all my other computers

• The agent transcribes all the videos in the footage archive whenever new footage is added

• I talk to the bot over Telegram and give it specific instructions eg “Hey I just added new footage for artist X, can you please come up with a couple story ideas around marketing videos for the footage we have?”

• It will then send me a couple draft videos - these first drafts tend to be quite unpolished and nowhere close to good but I can get a very good idea for what the story is

• I will work with it to iterate and polish the versions I like till it gives me something good eg. add in subtitles, replace specific scenes with other scenes, etc

• I initially let the agent use FFMpeg for video editing which just outputs a video file but have been experimenting with using hyperframes video editor (link).

• This has the very nice benefit of I can pick up editing exactly where the agent left off so it’s like collaborating with the agent on multiple projects in parallel.

• It can also generate copy for posting this onto social media which is super useful

Results

Here’s the first cut video:

It’s not perfect but it’s definitely a good enough jumping off point for manual editing after. I could keep talking to it to make it pixel perfect but I feel like it would be quicker to finish the last 20% manually.

My next step is to integrate into Da Vinci Resolve’s latest version (released 2 days ago) which has an MCP so the agent can use Resolve directly. I think Da Vinci is much nicer to use than Hyperframes so quite happy to have that option now!

Other Solutions

Before I made my own agent I decided to have a look at what’s available on the market. There were a bunch of very interesting video editing solutions (Descript is great, tried it for a month) but none of them have this workflow of brainstorming multiple concepts and editing them into multiple deliverables - they focus more on either generating AI videos or on using text as a way to drive the video editor. Premier Pro had an MCP that could be used with Claude but I prefer not to use Adobe out of principle so now that Da Vinci has an MCP - that’s the next step!

I’m sure this space will change very soon and hopefully we will have some better off-the-shelf solutions but in the meanwhile - always nice to build something that does exactly what you need!

• As someone who’d had to do their fair share of sourcing I realized I have a couple rule of thumbs and general idea of where I would look and what I would look for to find a decent set of suppliers. I’ve encoded this in a file as a set of guidelines for the agent to follow.

• I then give it a description of what I am looking for. Here as a trial project I’ve given it this prompt:

  • “I am looking for a 200mAh lithium polymer batteries can you please find me factories near Shenzhen”

• It then presents me with a PDF full of options. I’ve uploaded the PDF and a screenshot of one page as an example.

Shenzhen Gba Shortlist Public Redacted

2.56MB ∙ PDF file

Download

Download

_{I’ve run this through a filter to redact any sensitive information because I don’t want to mass upload other people’s contact info on my blog.}

• I then ask the agent to contact all of them (or eliminate some if I don’t like them) and get quotes

  • One feature that’s only partially implemented is responses - many vendors tend to ask similar follow-up questions so I have asked the agent to create a knowledge repository for each project which it can then use to draft replies to vendor questions. Right now I manually approve these replies just as a guardrail.

• I can then get a very good idea of what the market for this product. I’ve shown price vs volume here as a basic graph but once you have the data you can do more interesting things like lead time vs cost comparisons etc.

• The idea here isn’t to take human judgement out of the equation - at the end of the day finding a factory that is a good fit is about more than just numbers, but it does make finding potential candidates a lot easier.

• Ethics note: Because this agent can talk to humans and ask them to do real work I only do this if I know there is a real need and I am going to purchase.

  • This means the graph above is based on the initial search and not validated quotes at the end of the process - I don’t want to make these people do real work just for my blog post.

The Prompt

Here’s the input to the PCB design agent:

“Design a PCB that is powered by a coin cell and turns an indicator LED on when the ambient light levels fall low”

Pretty basic but still a fun enough way to test it out.

The PCB

I then sent the design files the agent gave me to LCSC for PCB fabrication and the results are below. Very surprisingly it worked!

There was one issue: it got the footprint wrong for the coin cell holder (classic human mistake to be fair). For now I just worked around that by using a copper sheet to make a temporary coin cell holder.

Video of the PCB in action below.

What’s next?

This seems cool enough that I will likely continue playing with this - I can think of many things that need improving here. Let’s see if I can get this to have better design skills because the PCB it designed right now works but looks and feels super amateur. Good start though!

I show the output from my first iteration of this agent below.

Workflow

Here’s what the agent does behind the scenes:

There are two interesting things here:

1. How do you let the LLM reason about the electronic design?
   As an engineer you tend to have different rules of thumb/intuition/gut feel and the LLM has some of that but it needs stronger guardrails. I believe the answer here is simulation. I’ve used ngspice.
   
   As a human I would rarely bother with simulation because it just takes too long for the work I do. But an AI agents time is less precious and so I have asked it to go through and simulate the circuit using spice before moving on to the next stage. This allows it to spend a bit more time doing the electronic engineering reasoning work and gives it a bit more information to work with to make good decisions.

2. How do you get the LLM to do PCB layout and routing? This may be controversial - use an autorouter. Freerouting is what I have used. Most humans prefer not to use one because the output is not great unless you set up a bunch of rules and by the time you’ve setup those rules as a human you may as well have manually routed it for smaller boards. But again this equation is not quite true for an LLM - the visual layout work is not easy for it but setting up autorouter rules is much quicker and easier so you add in an autorouter into the pipeline and ask the LLM to just setup the rules for it.

Ok cool but what do the boards actually look like? I ran it on two projects and I’ve posted the output below. I have also sent one of them off to get manufactured and will post the results when I receive them!

Project 1: Thermocouple to UART converter

This first project is what I used as a trial project to get the agent running so there were plenty of issues and it was just like any other software project where I iterate to get this working. I won’t go through the details here because it would take a bit too long to write and is less interesting as an example of what the capability of the tool is.

Prompt

“Design a K type thermocouple to UART PCB”

(yes that was really it, but then there was a bunch of debugging back and forth while I bring up the tool)

Output

Quite usable in my opinion. Shown below. Could be more space efficient, but other than that it seems perfectly usable.

Project 2: Design a low-light detector

I’ve attached screenshots of the conversation at the end of this post. Summary of what happened below.

Prompt

Design a PCB that is powered by a coin cell and turns an indicator LED on when the ambient light levels fall low

Issues

• It started with recommending a blinking LED to save power but it later realized the timer circuit that it built would take up too much power - this was fine since it flagged it to me I told it to not blink the LED.

• The schematic is really annoying to read. Technically fine but absolutely not how a human would write a schematic. This needs improvement.

• It initially forgot the copper pour - I reminded it and told it to add it to the workflow. Wouldn’t have affected functionality but definitely a miss.

• PCB size could be smaller - size optimization needs improvement.

And that was it.

Output

Loading the output from the agent into JLCPCB worked great and no issues here.

The only annoying this was I would have to pay $25 to assemble the coin cell holder and through hole parts so I decided to just solder those on myself.

Let’s see how it goes! I’ve uploaded the Gerber, BOM, and pick and place files in case anyone else wants to try it out: link to files.

Will update with results here :)

Conversation Record

Safety

My general approach with all AI agent workflows is they must run in their isolated containers - OpenClaw runs in a virtual machine on my NAS and it only has read-only access to things like my personal email.

There are very specific things I give it write access to and those are things that I know I can reverse. For example:

• My AI agent has it’s own folder on my cloud drive and I make daily backups of this folder so that if something goes wrong I know I can reverse it

• It has it’s own Obsidian Vault (my notes app) so it can interact with my notes but within it’s own sandbox

• Similarly I have given my AI assistant it’s own phone. I had an old android phone lying around which has now become my AI assistant’s phone. It was wiped clean before setup and now it only has access to the apps I choose to give it access to.

Phone Access Setup

Here’s how I setup phone access for my OpenClaw instance:

• Enable developer options on Android

• Enable wireless debugging

• Give your agent access to an adb skill and then start wireless pairing (link for reference: https://gadgetguideonline.com/android/android-adb-wireless-debugging/)

• Once you’ve done that you can have your agent use your phone!

• I have found that Wireless Debugging turns itself off occasionally so I am likely going to switch this over to USB debugging at some point.

The obvious question now is what is this going to be used for?

I have found a couple of interesting use cases here and will write about them over time but the general idea is that phones are a lot more trusted vs browsers. Agents using browsers keep running into captchas and other “are you human?” checks.

By giving my agent it’s own phone I can allow it to run more complex workflows without getting interrupted. It also turns out that phone apps condense information onto a screen quite well and this makes it easier for an AI agent to use.

I will write more about some of these use cases in the future.

• I give the tool a product idea

  • It performs market research and competitor research

  • It provides initial research into operational requirements (licensing requirements, any other operational concerns etc)

  • It generates branding (brand name, colour combination, logo, taglines)

  • It generates product concepts and images

  • It generates a landing page

  • It generates ad copy and visuals

While normally using the tool, after each stage it stops and asks for input on what to change before moving on to the next step. For now to demo the tool I’ve let it run on an example fully autonomously without my input (except 2 things - I’ll talk about those later).

I’ve added the output of the workflow below. I think the quality of the initial output is a 5/10 but I’m quite happy with this workflow as a first iteration and I think there’s easy ways to improve the quality massively:

1. Giving it feedback between each stage already improves the output quality significantly - it was an interesting experiment to run it on full autonomous mode to see the output + it made writing this blog post easier

2. I can likely also improve the prompts to give it more context and “training” (eg. feeding it a couple of articles on copywriting techniques before asking it to write copy)

3. Using specialized tools like Leonardo.ai API for specialized portions of the work like image generation (Leonardo Blueprints are amazing)

4. I also need to ask it to include more detailed citations for the the research bits (eg. competitor research) so that I can cross reference important figures later

I’m quite excited about the potential this type of workflow has on product development - being able to validate a product before building anything is going to get much much easier.

The Output

First, here is the full output: https://github.com/IcedRooibos/Mushroom-Chips-Validation-Workflow-Output-Example.

I will highlight some interesting bits below.

If you want to see the chat that led to the output, I’ve also attached a copy of the conversation at the end of the article - it’s quite cool to see it do its thing with minimal input.

Market Research and Competitor Research

• It’s tried to estimate market size (but only found global market size info which isn’t very useful in this context).

• It also found an interesting list of competitors to use as a starting point to further understand the market

Operational Considerations

It talks about import regulations and packaging requirements as well as estimates costs across different manufacturing models. I think the estimates are off but the line items being included are a good starting point to fill in better numbers.

Branding

It came up with a brand name and some logo options - I think both the brand name and logo are mediocre and this is likely where human input would have been very helpful.

Product Concepts and Images

As well as some product renders - but the branding isn’t consistent across those and the images aren’t great. This bit needs work and integration with Leonardo.ai will likely go a long way here.

Landing Page

It came up with a landing page. Link to the site: https://umamichips-bfppijdq.manus.space/

Images below in case I ever take the site down later.

Ad Copy

It came up with ad copy and images to go with those ads.

One variant here, but there’s a whole library of them - including platform specific ads eg. Facebook, Instagram, HKTVMall (local Hong Kong online shopping platform), Google, etc.

The Chat

I think it’s quite cool to see the whole chat play out so I’ve attached an image gallery here with the chat.

Human Input

After sending it the initial prompt I saw it was looking into manufacturing in Hong Kong which I knew is not a good path to go down so this stopped it and told it not to consider that as an option.

In a very human-like moment of forgetfulness it forgot to make the landing page so I reminded it to do that as well. It’s OK, AI, it’s only human to forget.

For 20 minutes and ~3.7USD worth of work this is a really interesting starting point. I am sure this will only get better.

A product validation approach I’ve been thinking about that uses AI to speed up the boring parts. The idea isn’t to remove the human from the loop—it’s to make the loop so much faster that it feels like a completely different process.

I think 10x faster/cheaper tends to have more of an impact on how you work than just “a bit faster.” When we moved from film photography to digital the workflow changed. With film you were more careful about what you shot, you had 36 shots in a roll. With digital you click as many pictures as you want and delete the ones you don’t need later. When something takes a weekend instead of a month, you’re more willing to try it. You’re more willing to throw it away if it doesn’t work. That’s the goal here. (Philosophically, I think it’s an open debate as to which approach is better but certainly the second approach wasn’t possible before so I am interested in exploring it).

I haven’t battle-tested this yet so take it with a grain of salt—but I think the approach is interesting enough to write about now and then follow up with results later after I have run through this a couple times.

The Three Phases

I’ve broken this down into three chunks: Initial Research, Validation Marketing, Iteration.

Here are the goals:

• Initial Research: This is for you to look into the idea and learn about the lay of the land and decide if you want to take this to the next step

• Validation Marketing: I call this validation marketing because the goal isn’t to sell but to learn what works and collect an audience along the way (e-mail lists, followers, etc.)

• Iteration: This is the human in the loop - take the learnings from the previous two stages and decide what to do next. Tweak the product concept? Change the ads? Start over with another product concept entirely? If done right, you now have an audience from step two to talk to about your tweaks before you start the loop again.

A Quick Note About Tools

Here are the tools I find useful and have used for items in the rest of the post below.

(Note: I gain nothing from you signing up to them, this is just to share my current workflow).

• Kagi - privacy friendly search engine

  • Their search-assisted AI is the best I have found so far. I use the Ultimate Plan as it has a special research mode.

• OpenWebUI - self hosted chatGPT UI

  • This is the “heart” of the system and how the output from a lot of these sections talk to each other. I add output from all AI models to an openWebUI folder (called a “knowledge base”) that the model can reference later

• Canva

  • Logo generation

• Leonardo.ai

  • Image and video AI models - I really like the outputs from this tool

• Huemint - colour palette generator

  • This comes up with colours palettes very quickly and I can cycle through a few before I hit something I think is a good fit for a specific idea

• Carrd / GetResponse:

  • For landing pages.

  • I have used Carrd for a long time and it’s a good tool

  • GetResponse claims to be more integrated from an analytics and email marketing perspective. It even offers basic ad management. Seems interesting enough that I am taking it for a spin.

Initial Research

The goal here is to quickly figure out:

1. Are people talking about this problem? — Find mentions of the problem or similar problems online. Forums, Reddit, Twitter, product reviews complaining about something. You want to know if real people actually care.

2. What competitor products exist? — Who’s already selling something in this space? What do their products look like? What do customers say about them?

3. What competing solutions exist for the same problem? — Sometimes the competition isn’t a direct competitor—it’s a workaround, a DIY solution, or just people tolerating the problem.

This is also a good place to add constraints that are important to you (for example “Could this go from idea to revenue in 90 days?”).

How?

• Kagi’s research assistant is great for the research phase to try and find information and put together a summary to the questions you feel are important to have answered

• Once I have done AI-assisted research I will move the collected information to OpenWebUI so that future conversations can reference this research

The output of this phase: enough context to decide whether to keep going, and enough material to inform your product positioning.

Marketing

Once you’ve decided the idea is worth testing, you need to get something in front of people. The goal is minimum viable marketing—just enough to run a credible test.

Minimum Viable Branding

You don’t need a brand agency. You need:

• A brand name (openWebUI brainstorm with AI with previous conversation context, pick one)

• A tagline (same)

• A color palette (I use Huemint)

• A logo (Canva’s AI logo generator works fine)

This takes maybe 15 minutes. It won’t win design awards but it’s good enough to pass the sniff test and get the audience to look at the product and decide if they want to demonstrate interest (eg. by signing up to your e-mail list).

Audience Definition

Who exactly are you selling to? I work through this with AI as well:

• What’s the problem? What are current alternatives? Why are they inadequate?

• Come up with 3-5 candidate customer profiles

• Pick one primary target (you should do this one, not AI)

• Write a one-sentence customer definition

• Sanity check: are they reachable? Is the market large enough? Use your advertising platform of choice’s targeting tools to check.

In reality I do this before branding because this should feed into the branding but to read a blog post it flows better this way so I have left branding first.

Asset Generation

This is where AI image/video generation is genuinely useful. You can create product visuals without having a product.

• Pull reference images from competitors or suppliers

• Use AI tools to generate product shots, carousel images, lifestyle images, with the branding you want

• Generate short videos from images if needed

I’ve been using Leonardo.ai and the quality is good enough for ads and landing pages.

It’s quite believable that AI could do marketing research and come up with some audience definition information or landing page copy. It’s possibly a bit harder to believe that it could do it well.

But asset generation seems like the kind of thing AI can’t get right. It was the missing piece for a while, in my opinion. But the models are good enough now to get what we need out of them.

As an example, if this is the input to the system:

Here’s what we can get out of it.

Video for advertisement:

Image carousel for landing page:

Copy Generation

• Landing page copy: I feed the knowledge base context to openWebUI brainstorm with Opus to write landing page copy

  • I have a skeleton structure for the landing page that I feed to the AI as well so it knows what kind of things to generate. In theory we are leaving some creativity on the table in terms of structure but a well formatted landing page is a good starting point and my hope is the speed we get out of this will make it worth it.

  • Here’s an image of the prompt I use (+ it has all the knowledge of conversations prior to this):

• Ad copy: Same process, but formatted for ads

• Ad scripts: If you’re doing video ads, get AI to draft scripts based on your audience and positioning then generate then using your tool of choice

Landing Page

I have made a template in two website builders based on the structure I shared above and can plug in the copy and assets. Nothing fancy. The point is to have somewhere to send traffic to.

In terms of tools I am currently testing both Carrd and GetResponse. I have used Carrd for a long time and it’s solid, but GetResponse claims better integration with analytics and email so I am taking it for a spin. I’m quite happy to fall back to Carrd if GetResponse causes issues. This is an implementation detail anyway and not really core to the experiment.

Iteration: Human in the Loop

At this point you’re ready to run ads and start learning. The human is still very much in the loop. You’re making decisions about what to test, how to interpret results, whether to iterate or kill the idea.

The difference is the loop is now a lot faster and cheaper to execute. You can test an idea in a day instead of a couple weeks. You can test multiple ideas in parallel. And when something doesn’t work, you haven’t sunk much into it. This also sounds like the type of process that is easy to operationalize and scale; run by a small team testing multiple ideas in parallel. Something to explore later.

What’s Next?

I’m going to run a few ideas through this process and see what happens. Will report back on what worked and what didn’t. I think there are some parts of this process that raise a few questions (how do I know if it’s a product issue or a ad targetting issue, for example?) but I am hoping as I run this playbook a bit more I can learn about what happens in the real world.

If you have thoughts or have tried something similar, I’d be curious to hear about it.

For example, if I ask it to “find me a cheap lithium battery charging integrated circuit,” it will either give me an answer that is outdated because its training data cut-off is in 2021:

Or, if I enable the web search plugin, it will give me a very brief summary of the top five links it finds:

The answer looks useful, but it has not really done the research. It has given me a few places to begin doing the research myself.

However, I think this is a bit counter to how we should be using computers - computers are supposed to go crunch the numbers and figure stuff out while we are at the beach sipping margaritas. The reason ChatGPT needs to respond quickly is because of our culture of instant gratification, but this is counterproductive.

Instead, I wanted to develop a tool that would take its own time, do a fair bit of work for me, and then let me review its results as a starting point for my research.

How it works

After playing around for a weekend, this is the structure I came up with:

• Enter an input query

• Use the OpenAI API to generate search terms

• Ask the LLM which websites should specifically be checked

• Get the results for those search terms

• Read through each result and extract any relevant information

• Produce a summary of the relevant information on each page

• Use all of those summaries to generate an overview

The user interface is good old email.

I send a question to the tool and it replies with the completed research. This means it can take as long as it needs without making me sit in front of a chat window and wait. People don’t expect emails to get instant replies so it works out from a user expectation perspective.

Trying it on real questions

Finding a battery charging IC

One of my test queries was:

Find me the cheapest battery charging integrated circuit for a single-cell lithium polymer battery.

The tool generated search terms, read through the results, produced an overall answer, and included a summary of every page it used. I also added a “lateral thinking” section that tries to reframe the question in case the original approach is wrong.

It does a lot of this asynchronously. If there are 50 search results, it gets the content for all of them in parallel and extracts the relevant information in parallel. Even then, because of API rate limits, it can be a pretty slow process - a deep query can easily take 15 to 20 minutes. But thinking about what is actually happening in that time, the LLM is doing the work I would have to do anyway.

Here is the full email it returned:

The result is not perfect. In fact, it failed to answer the question: it did not find a specific IC and compare actual prices. But it did collect much more context than the normal web search response, and the per-page summaries make it easier to see why the final answer is good or bad.

That is important because I do not want the output to be treated as a definitive answer. I want it to be a useful starting point that I can verify.

Slow does not automatically mean good

The biggest lesson so far is that giving the model more time and more information does not automatically produce a better answer.

I made “standard” and “deep” search modes, with options to either discard irrelevant pages or summarise everything. Surprisingly, deep search was sometimes worse than standard search. More sources also means more irrelevant information, more opportunities for a bad summary, and more confident nonsense in the final answer.

I found some particularly strange hallucinations about air pollution and climate change in completely unrelated research. After digging through the logs, I found that some summarisation prompts were being sent with no content to summarize. The model happily wrote a plausible summary anyway. I now ask it to respond with “Content missing” when there is nothing to summarise.

There is also a cost and processing-time problem. One experiment made 56 calls just to clean up individual source summaries and another 324 calls to improve partial answers:

A particularly deep query ran from around 11:07 p.m. until 2:03 a.m. and cost almost US$2. I cut that to about US$0.20 by skipping calls when the input text was too short, but the result was also garbage.

I also tried moving more of the work locally using Hugging Face summarisation and question-answering models. Facebook’s BART CNN model and Google’s Pegasus reduced the amount of text that needed to go through ChatGPT, but the quality was inconsistent. A simpler prompt made Pegasus much better, although there is still plenty of work to do.

This configuration - the processing pipeline - is just what seemed reasonable after a weekend of experimenting. I am sure there are better ways to do it.

A more directed approach

One alternative is a ReAct-style model, where the LLM researches sequentially:

Question: Find me a battery charging IC
Thought: What battery charging ICs exist?
Action: Search for battery charging ICs
Response: TP4056, BQ24075, etc.
Thought: Which one is cheapest?
Action: Search for the price of each candidate

Instead of obtaining all the information at once, processing all of it upfront, and then churning out an answer, this would produce more directed research. The primary advantage is that it should be cheaper and use fewer API calls.

What happens next?

I have not worked on this much beyond the experiments above. The processing time is long, the results are not always accurate, and there still are not enough data sources for some of the questions I care about.

More importantly, combining search with an LLM feels like an obvious native feature for search engines. It seems likely that they will integrate this directly, and I do not see much point in competing with that.

For now I am leaving AutoResearch here as an experiment.

I still think the basic idea is interesting: an AI does not always need to answer instantly. If it can spend 20 minutes doing work that would take me an hour, that seems like a very reasonable trade.