Independent software
Software that
answers to you.
Not to advertisers, not to a monopoly, not to a growth chart. We build for problems we see every day, with privacy and security from the first commit.
§1 Why we build
The tools we depend on
shouldn't own us.
We want many independent makers building the software people rely on, so no single company gets the final say over how we live and work. Our part: excellent software for real problems we see, private by default and secure from the first commit. Before we build anything, these have to be true.
- A real problem
- Something people deal with every week, not a demo looking for a use.
- A way to know it worked
- If we can't say what better looks like, we don't start.
- A difference we can make
- Plenty of good ideas are someone else's to build. We pick the ones we can do well.
- Something we'd still be proud of
- Years from now: still running, still trusted, still worth opening.
§2 How we work
Built faster with AI.
Made right by people.
AI made code cheap to write. It didn't make it right. Agents do the heavy lifting; people with the experience decide what to build, which trade-offs to make and when it's good enough. Nothing reaches anyone until it's tested, reviewed and approved.
Who does what
- 01
Intent
People
We set the goal: who it's for, what it has to do, what must never go wrong, and how we'll know it worked.
- 02
Plan
People
We break the work into small steps, weigh the trade-offs and decide what's worth building first. AI helps us explore; we choose.
- 03
Build
Agents, steered by engineers
Agents write the code with the right tool for the job: native on the phone, Rails on the server, on-device models when data shouldn't travel.
- 04
Prove
People, with tools
We test, measure and check it's correct. A second model reviews every change and risky ones get a security review. Fail any of it and nothing ships.
- 05
Run
People
Deploys without downtime, rollback, backups and isolation, planned before anything launches.
A person approves every release
The bar
- Correct
- We test what it must do and what it must never do. People review the results and decide when it's ready.
- Available
- We plan how each release goes out, watch it in use, and know how to take it back.
- Resilient
- We design the core to keep working when a provider fails, AI providers included. AI features can step aside; your work shouldn't stop.
- Secure
- Threat model first, least privilege, encryption by default, customers kept apart, and as little personal data as possible.
- Right tool
- No house stack. Swift on an iPhone, Rails on a server, a model on the device: whatever the task needs.
- Results
- We measure what changes for the people using it. Lines of code don't count.
A real change
- intent
- A photo must never be deleted by mistake.
- plan
- 6 rules: re-check every batch just before deleting, any face protects a photo, all or nothing…
- build
- 12 files changed · +289 −88
- prove
- 127 tests · deletion suite extended
- record
- commit 89fe9f2
§3 The platform
Every app
makes the next one better.
We didn't want to start from zero each time, so we built the machinery once: the gates every change passes, the way releases go out and come back, the hosting every app shares. Each app we ship leaves it a little stronger for the next one.
- One road to release
- Every app ships through the same gates: tests, a review by a second model, a security review for risky changes, and a person's approval.
- Agents with narrow permissions
- They can propose and build. They can't ship on their own.
- Releases you can undo
- On our servers, a release goes live in one step and comes back the same way.
- Each app at its own address
- Apps live under devserker.com, on shared, hardened hosting, kept apart from one another.
§4 Principles
More capability.
The same responsibility.
People decide.
We set the goals, plan the work and choose the trade-offs. AI helps us think and build. A person approves anything that matters, and answers for it.
Proof over confidence.
Generated code gets tests, an independent review and real measurements. “Looks right” is not a test.
Security is the floor, not a feature.
It isn't on a pricing tier and it isn't a later milestone. It's in the first commit.
Less data, kept private.
We collect as little as we can, keep it on your device where practical, and never sell it.
Leaving should be easy.
We treat taking your data with you as a requirement. Nobody should stay with us because leaving is hard.
Results, not output.
Agents can produce endless code. We count problems solved.
Respect attention.
Reminders stay optional and nothing is built to hook you. Good software helps you finish what you came to do, then lets you go.
§5 Contact
Curious what we're making next? hello@devserker.com