The software house is dying.
Not the people in it. Not software. The business model: a client describes what they want, a team builds it, the team bills for the hours, the code gets handed over, everyone moves on.
That model is collapsing, and we don't think it comes back. It's also why, as of October 1, 2026, we have paused new custom software development at Bolder.
What actually died
Start with what a software house used to sell.
Hours. A pool of developers, billed by the day, writing the code a client couldn't write themselves. Most of that code was never hard. Login flows. Admin panels. CRUD endpoints. The tenth version of the same dashboard.
An AI coding agent now writes that code in minutes. Your client knows it. Their CTO has watched it happen.
So when a software house bills two days for an endpoint, one of two things is true. Either the team is not using the tools, and the client is paying for the slow way. Or the team is using them, and billing for time it didn't spend. Neither one survives a client who asks the question.
The cheap MVP went the same way. A generic first version of an app used to be a real project. Today a founder with a weekend and an agent can get most of the way there. Undifferentiated software is a commodity, and commodities don't pay for an office full of developers.
What's left
Three things still need people who know what they are doing.
Orchestration. Someone has to frame the problem, split it up, direct the agents, review what they produce and decide what ships. That is a senior job. It doesn't scale with headcount.
Domain depth. Grading fruit on a production line, reviewing contracts in two languages, running a sales agent that handles a wholesaler's out-of-stock items and changed orders. The code is the easy part. Knowing what "correct" means in that domain is the work.
Owning the outcome. Someone has to be on the hook when it breaks in month three. Not just for the build, for the result.
Here's the shift in one table:
| The old software house | What replaces it | |
|---|---|---|
| Team | Large pools of junior and mid-level developers | Small teams of senior operators |
| Pricing | Billable hours and headcount | Outcomes and risk taken off the client |
| Core skill | Writing code by hand | Framing problems and directing AI |
| After launch | Hand over and move on | Own it in production |
That last row is where the build-for-hire model breaks for good.
You can't own the outcome of software you handed over.
Where that left us
We have done both kinds of work. We built for clients: a contract review system for a Saudi enterprise, a computer vision line that grades dates, a sales agent with 20+ tools for a food and beverage wholesaler. We also built our own products, bolder.fit and Scailor.
Look at what compounded.
The client projects ended at handover. Good work, real results, and then the code lived somewhere we no longer touched.
The things we kept got better every month. The agent harness we built for one product now runs in two. The evaluation engine we built to check our own AI became a service. Every lesson stayed in the building.
If the value of a software team is now in operating systems over time, then the team should operate its own.
What we're doing
As of October 1, 2026, we are pausing new requests for custom software development.
Until June 2027, we will only take on Building Blocks™ requests: integrations and builds based on components we already run in our own products. This is our Focus mission. It is also why we sunset the BEval Studio platform and kept evaluation as a service.
| Building Block | What it does | Used in |
|---|---|---|
| Bolder Agent | Our tool-calling agent harness. It connects to your app's tools over MCP, asks a person before risky actions and tracks the cost of every step. | Scailor, bolder.fit |
| Bolder AI Food Logger | Turns a meal photo into a food log, with portion estimates and a confirmation step before anything is saved. | bolder.fit |
| BEval Judge Workflow | Evaluation as a service: deterministic verifiers, LLM judges and rubrics written for your product. | BEval Studio |
| Gamification Pack | An ebook with our research, design spec and lessons on streaks, badges, points and leaderboards. | bolder.fit |
Anything we build for someone else now has to be something we already run, and something that gets better because we run it.
If what you need fits one of these, book a call. If you have an ongoing project with us, we will contact you directly about what this means for it. Anything else goes to support@bolder.services.
Maybe the software house finds a second life. Some will, as small senior teams that own what they ship.
We'd rather own the products.
