Devtools - what are you seeing in early-2023?

Devtools - what are you seeing in early-2023?

Running Console means I see a lot of tools. From open source projects through to small terminal utilities and large public cloud providers, I play around with a lot of devtools, cloud services, open source, and anything developer-first. I’m interested in anything where developers are the primary user.

I’ve also been investing in developer-first startups for about 18 months. I like to get involved at the early stages (primarily pre-seed and seed, but also a few series A deals), typically investing $25k. I have invested in 27 startups so far. My criteria for investing are the same as the selection criteria for inclusion in the Console newsletter, but with some added questions about the team, timing, flow, and whether the idea is truly venture-scale.

With a broad view of what’s going on, whenever I meet someone I’m always asked what are you seeing? What is the current state of devtools? What is interesting? So because I like to write to solidify my own thoughts, here are a few things I’ve noticed recently:

Developers love simplicity, but with a power-user mode

Five years ago, if you’d asked me whether there was an opportunity to build another large-scale platform for compute then I’d have said no way. AWS has a compute product for everything. Whether it’s EC2 directly or other more PaaS-like products such as Fargate and Lightsail, if you need to run your app code then AWS has an option. And if not, then how about GCP or Azure or DigitalOcean or Linode?

So I would not have predicted the rise of alternative compute platforms like Vercel, Netlify, Railway (I’m an investor), and Fly. Why have they been so successful?

Simplicity.

A common theme amongst the products that do well in the Console newsletter is that they make annoying, tricky, and complex tasks much easier. For developers, that means getting their code into production and dealing with a lot of the scaling issues with no ops required. Vercel, Netlify, Railway, and Fly do a lot of things, but they all start by building and deploying your code to the cloud with minimal (or no) changes.

But simplicity is only one requirement, it’s not sufficient. I see a lot of new cloud platform startups, particularly those which claim to abstract the complexity of the cloud, but they tend to restrict rather than enable. Simplicity is fine for “hello world”, but real apps need more. Logging, metrics, access control, storage, databases, regions, serverless functions…there are a lot of things that cloud platforms need to provide if they are to be anything other than toys for side projects.

It’s great to be able to deploy your code within a few seconds without having to think about containers and servers and OS updates. But the platform also needs to capable of providing advanced functionality. It’s not going to work if developers have to go to a public cloud provider once they “level-up” past a prototype. There needs to be an upgrade path from toy to prototype to MVP to scale. This applies not just to cloud services, but to all developer products. This is simplicity, but with a power-user mode.

However, it can go too far - I’ve seen that some tools can be too complex. Subscribers of the Console newsletter are not particularly interested in Kubernetes or GraphQL, for example. Both of these get less engagement when we feature those types of tools, and GraphQL is particularly disliked. The problem with these is power user mode is the default rather than an option. They’re too complex for most developers.

That presents an opportunity. I’m an investor in Grafbase which aims to simplify GraphQL. I also invested in Railway which abstracts away all the hassle of running a k8s cluster by taking your code (or containers) and deploying it for you. Companies like Stellate are making it easier to deploy and scale GraphQL APIs. The power is still there, but it’s behind opinionated defaults, easy onboarding, and a great developer experience.

DALL-E: "a futuristic computer flying through the night sky in digital art neon style"

General purpose database first, then specialize

Postgres is the ultimate general purpose database, especially the more recent versions which support functionality that was once only found in specialist databases e.g. JSON queries or key/value types. There’s a reason why AWS Aurora and Google AlloyDB are Postgres compatible - it’s the default database for most developers.

But as workloads scale, it makes sense to reach for a specialist database. I’m constantly surprised at how much room there is for new databases. DuckDB and ClickHouse are examples of recently funded startups behind the open source projects, there always seems to be something new for SQLite (rqlite, LiteFS, Dqlite, WASM), and even innovation coming on top of well-established databases like Planetscale (MySQL) and Neon (Postgres).

Yet unlike many other developer products, databases have decades of examples of successful monetization. We know companies will pay for databases (licenses, support, services, tooling, cloud services, backups, monitoring, analytics), so perhaps the abundance of database startups shouldn’t be so surprising.

DALL-E: "renaissance landscape with an computer software company office building"

No/low-code tools to save developer time, but there are too many

We use Retool at Console (they’ve also sponsored the newsletter). It saved a huge amount of time building the admin panel for our jobs product and it’s amazing how flexible and customizable it is.

If you build a table of data you know the first feature requests are going to be pagination, sorting, filtering, searching…all tedious to build, especially if it’s for internal use rather than delivery user benefits. Retool’s table component gives you all of those for free. The value is clear - use Retool where it doesn’t make sense to dedicate developer time to building undifferentiated internal tools, processes, scripts, apps, and systems.

Airplane is another interesting example (also a Console sponsor) focused more on developer flows like executing scripts, database queries, and other workflow actions that would otherwise need ops to log into production systems.

However, the success of Retool has spawned hundreds of copy-cats. Some of these are doing something different, such as Appsmith being entirely open-source, but there are now too many. There is clearly a lot of demand for tools that allow non-technical users to create systems that would’ve required lots of developer time in the past, but how many can the market support?

DALL-E: "lots of clocks falling into a vast ocean on a planet with stars in the background"

Javascript & Typescript ecosystems get all the hype, but there are many other opportunities

If you agree that there are too many no/low-code tools then what do you think about the number of frontend frameworks!?

“Fullstack” used to mean you could build with both frontend and backend technologies, but these days it really means you write everything in Javascript (hopefully Typescript) first. That’s why there are so many interesting tools for that ecosystem - from the likes of Turbobuild to Socket and from React Email to Inngest.

However, whilst the Javascript & Typescript ecosystems get all the attention, there are still huge opportunities elsewhere. Python is particularly relevant due to its use in data science, machine learning, and AI. Rust is popular for environments that demand high reliability, performance, and security. Go is good for network and API programming.

These are all areas with lots of money available to buy products for mission-critical workflows.

DALL-E: "a crowd of people looking at a big screen with several other screens at the side and only a few people looking at them"

Developer adoption, company purchase

One of my favorite product-led growth examples is how Trunk (I’m an investor) have designed their product for developer adoption, but company purchase. The CLI is free and helps developers manage their code quality formatting and linting tools. It provides real value via an IDE plugin and pre-commit hook to ensure your code follows formatting rules, applying auto-fixes where possible.

However, it’s just the first layer in a suite of tools that become useful in a team. Trunk Merge helps coordinate and land many pull requests to avoid conflicts and waiting on slow builds. Trunk CI Analytics provides insight into each build step, highlighting where the most time is spent or errors regularly occur. These are trivial to enable if you’re already using the CLI.

Semgrep is another example where the CLI provides immediate value to developers, but teams can pay for rule management and vulnerability detection features. It can also be managed by Trunk.

Whenever I start a new code project, the first thing I do is git init. The second thing is trunk init. What comes next?

DALL-E: "a man in a suit buying items in a store pixel art"

Anything else?

AI is the obvious omission from the above, I made up for it by using DALL-E generated images because I don’t have much to add. GitHub Copilot is the easiest $10/m to justify with the improved productivity it gives me. It will be interesting philosophically to see the outcome of the lawsuits against GitHub, but it’s too late.

I’m also interested in the infrastructure layer, like Replicate and Lambda, because tools for builders are the best way to commercialize the early days of any new technological trend. I’ve not seen any other interesting applications of AI for developers, yet. I expect that to change.

Developer-first security is another area to watch. The vendor space is very fragmented with no clear leaders (Snyk stands out), which is strange when you compare it to devtools and infra. Everything is an enterprise sale. Security requirements are not going away, so why isn’t it easier for developers to build secure applications?

We’re still at the beginning of the age of the developer, so if you’re building something developer-first, reach out!