The Gumloop Handbook

👋

Here's a walkthrough of everything we as a company believe in, why we're building what we're building and where we hope to end up.

Why does Gumloop exist

We want anyone to be able to automate their work with AI. We started Gumloop in April of 2023 and have been working on the same problem since day 1.

Origin story

Gumloop was originally called AgentHub and Version 1 was a simple UI wrapper around an autonomous agent framework called AutoGPT. At the time, to use the framework you needed to be somewhat technical because you had to clone the code and set it up locally. We built a UI for fun and hosted the framework in the cloud so that all of those non-technical people could get to play with agents as well.

Agents felt like the first time non-technical people could really tinker with AI and apply it to their own use cases.

The only issue with agents was that they involved AI in every step of the process. This resulted in major issues with reliability and cost efficiency.

We would see the sort of tasks people were begging the agents to accomplish and realized that AI agents were normally overkill. For most of the tasks we realized that AI was only needed at key decision points and that most of this work, some python scripting, a few API calls and some AI here and there would perfectly solve the problem.

We started building an automation framework as a way to give non-technical people the ability to build their work but in a more cost effective, steerable and reliable way. It grew organically, feature by feature, into what it is today.

Want to see the world's craziest coincidence?

The first time Rahul (co-founder/CTO) saw the product was captured on video. If you skip to 4:59 in the video, you'll hear me say. "My co-founder quit this morning so I'm looking for a new one" (that did actually happen, he quit and said AgentHub wouldn't work at 11 AM that day). The next person to speak in the video is Rahul, excitedly asking a question about the product. He was contributing code to the project 24 hours later.

(I can barely watch this video back because of how monotone and boring it is. They asked if anyone else wanted to present, I raised my hand and winged it because I didn't have much to lose)

To read a bit more about what the product looked like on day one, check out this blog post.

What is Gumloop today

Gumloop is a platform for automating complex work with AI and a simple drag and drop interface. We let you build really powerful workflow automations that use AI without needing to hire an engineering team or be an engineer yourself.

What do we want to achieve

Let anyone have the impact of an engineer

We want to enable every non technical person in a company to solve their own problems without needing to loop in an engineering team. Anyone should be able to deploy their own AI powered tools that 10x their teams productivity without needing to write a large spec, sync with leadership, allocate budget etc.

Understanding the task should be the only requirement for automating it in the future, no other expertise required. At the moment, there is a bit of a learning curve to the platform but we're working on that. (See "where we're going"). Gumloop is currently perfect for 'semi technical people and above' but soon it'll be the go-to business automation engine for absolutely everyone.

Get our users promoted

Almost every existing business has massive inefficiencies. We notice Gumloop starts to spread internally within a company when a "champion" starts automating work and showing their colleagues.

This champion always has the same 3 characteristics:

1. They understand the problem(s)
2. They know what AI can do
3. They know Gumloop exists

If you're reading this post you probably have #3 locked down (or you're super lost). 1 and 2 tend to come naturally to that person because they spend time around the problem as a result of their job and they're excited by AI enough to frequently use LLMs.

This person will often become the "AI person" internally and will make their leadership team shed a tear of joy because they're finally "leveraging AI" like the CEO asked, but in a useful way.

We want to be their secret weapon that makes their division 10x more efficient and gets them or their team promoted.

Edit: this secret weapon phenomenon actually happened with one of our first enterprise customers.

Where We're Going

We want Gumloop to be such an outstanding product that marketing is an afterthought. Word of mouth from thrilled users should be our main growth engine. To get there, we have some core improvements to make to the platform that we we're building towards

Reducing the learning curve

The learning curve of the product is too steep.

Our power users absolutely love the product. Many are spending upwards of 5 hours a day building on Gumloop. The main problem is that the majority of people who hear about us never get to see what our power users see. They often drop off before they get that 'aha' moment and start building automating every aspect of their work.

If we can make that first 20 minutes on the platform less confusing and handhold users towards their first useful automation run, we'll have solved 80% of our problems.

This is fundamentally a UX problem that will take tons of iterations, creativity and native AI features. We're working on several solutions to this right now but eagerly hiring a founding engineer to help accelerate. See the job posting here.

Adding integrations

Every new integration we has a massive impact on the number of possible automations that can be built due to the number of node combinations. Users consistently request new features and integrations and we consistently build them. We will continue doing so until you can literally automate anything.

Edit: We're letting you build your own nodes now on top of adding integrations regularly!

Improving UX

There are tons of ways we can improve the overall user experience. Things like creating complex nested flows, collaborating across teams and new node discoverability could be so much better for our existing user base.

There are also fundamental backend improvements we can do to make everything faster, even more scalable and powerful. For example, we recently added parallel step execution which made large automations up to 32x faster. We have several more core platform improvements we plan on implementing like this.

How We Make Users Happy

One key differentiator between us and competitors is how much effort we put into caring for our users.

Support

Gumloop can be used in an infinite number of ways. That's great but it also means our users can get stuck, confused or frustrated in an infinite number of ways.

Having absolutely top tier support plays a key part in ensuring our early users have a great experience. We take great pride in it as well.

We're quick to answer emails, we're active in our discord community answering questions and each of our pro tier users has a dedicated slack channel for 1:1 support. Being able to go back and forth with people actively using our product helps us understand their pain points and ensures they never get permanently stuck.

Same day shipping

We used to say we were like Amazon because we had "same day shipping". When a customer would a new integration or feature we'd try to get it out by that night's release.

Obviously this velocity got a bit harder to maintain as our user base grew but we still try to build customer feature requests as quickly as we possibly can.

Users seeing features they requests in the product within days helps prove we're truly building for them. They can participate in steering product direction by just asking.

Listening to feedback and iterating constantly

Every time a user complains about anything at all, we write it down and try to act on it somehow. The first version of our automation builder was an unusable hunk of garbage. Gumloop has only become what it is today because of thousands of small iterations that were driven by direct user feedback.

Sometimes this is super straight forward when people message us saying "here's some feedback". Most of the time it comes in unexpected ways. Every discord question, unexpected error or confused clicks during a demo call is feedback in it's own right.

We try to keep our iteration cycles extremely short so tweaks to the product can be felt continuously. Most of the time we're shipping updates nightly to address the days feedback!

Strive to feel embarrassed

Looking back at demo videos from even a month ago should feel embarrassing. If we're not embarrassed, the product isn't improving fast enough and we're failing our users.

Who We Want To Work With

At the moment we're a team of 2. We want to grow but also want to make sure the right people join the team. We're aiming to keep the talent density as high as possible and have put a lot of thought into what kind of person we're looking for!

Relentlessly resourceful

This is by far the most important quality we look for. This is also the most important quality in a founder which isn't a coincidence.

Being able to just make things happen. Digging deeper into the problem when you get stuck is ideal.

We as founders don't have a plan b, or an engineering manager to turn to when we get stuck. Obviously we collaborate and will be extremely supportive but having that same end-of-the-line approach to hard problems is a must.

Optimistic by default

Building a startup takes a lot of optimism. There are so many people along the way pointing out every reason why something won't work. Those people are boring. If we'd have listened to them, we'd have given up day 1.

We want someone who's looking for the reasons things will work and leaning into them. Alternatively, if something isn't working we want to find a new approach that does rather than getting discouraged.

Excited to build

When working as part of a tight-knit team, attitude matters a ton. The mood in the room is an average of those who fill it.

We want to work with someone who's as excited to build for our users as we are. Building as a team is really fun and we hope everyone we hire will add to that atmosphere, maintaining or bringing the room's average.

This screenshot above was from when we started Gumloop. We were working together 24/7 since this exchange.

There's a specific feeling you get when you meet someone who's excited about the same things you are. You leave the conversation with more energy than you entered it with. We're excited to build great product for our users. We hope whoever we hire is just as excited.

Reliable

We want someone we can rely on. You'll always be able to rely on us, but it'll need to be mutual.

Knowing that great work is being done without needing to second guess things is the dream.

Please reach out

If you think you fit these criteria and want to join the team please email me at max@gumloop.com

PS: One way to really set yourself apart is to actually try the product thoroughly. 99% of people who apply have clearly never tried to use Gumloop. If you do, you'll have a massive advantage.

What It's Like to Work at Gumloop

Here's what'd it'd be like to work on the team.

20 minute standup

We don't have 'sprints' in the standard sense. We find it impossible to plan 2 weeks in advance if we're updating the product daily. We have a 20 minute meeting each morning in which we discuss priorities for the day, almost like a mini sprint planning.

We try to keep all other meetings to an absolute minimum. If you need to talk to someone just walk up to them in person to chat.

Focused work

The majority of everyone's day is dedicated to focused work. That doesn't always mean working in a silo, it can be collaborative because in our view focused work is any time you're 100% focused on making the product better.

Sometimes we'll have a 2 hour whiteboard session working through design challenges. Sometimes we're fixing bugs or building completely new features. It all counts. As long as most of the day is spent doing this we're doing the right thing.

Papercut friday

Throughout the week we find many small issues with the product that aren't worth fixing immediately . We call these "papercuts". They're not anything urgent but they can add up and really ruin a users experience on the platform. On Friday we take the entire day to tackle those. This helps avoid easily fixable issues falling to the 'low priority' section of the backlog to gather dust.

Merging into production on Friday should feel like a breath of fresh air if we did this right.

Backlog grooming

We currently don't have any predefined sprints. We like to build as quickly as possible, iterate often and ship updates on a daily basis. New things pop up every day so planning more than a week into the future would feel silly with the current pace.

We have an ever changing backlog of tasks separated by priority and size. Urgent are the tasks we're focusing on today, High are those we want to tackle this week, medium in the next two weeks etc.

Everyone is responsible for adding tasks to the backlog and categorizing them to help define direction.

Offsite hackathons

Renting an Airbnb and having a company hackathons has resulted in some of our most impactful work as a company. No distractions. Just a bunch of people who love building cool product in one place working together.

We also try to take time to decompress when we're there and go outside. Last hackathon was in Whistler BC Canada and we spent an afternoon rock climbing. Spending some time away completely detached from work is important to the fun of it all.

Onsite hackathons

Sometimes we'll agree to collectively tackle a part of the product that needs love. We'll find a nice place in the office to set up, order a bunch of food, brainstorm on a whiteboard and start hacking into the night as a team.

Here's a timelapse of our last internal hackathon where we focused on UI details in the flow builder that had been bothering us. We revamped node styling, added an extra settings panel, added logos to the LLM selection menu, redid notes completely, added shortcuts for everything, added animated edges and fixed fundamental scalability problems with concurrent runs. We shipped all of this in 6 hours.

Moving faster and caring more is how we win.

More to come

If you made it all the way to this last section, you might be interested in joining the team!

If you're serious about it, send me an email at max@gumloop.com. I do get a ton of spam so please try out the product before emailing. I reply to everything that someone put effort into sending.

I'll keep updating this handbook as we grow.

- Max