For the first time in history, a practitioner without an engineering background can build a working digital product in a weekend. The implications for this sector, and for the communities it serves, are worth taking seriously.
A Personal Starting Point
A few months ago, I built a video game for my nephews. Not downloaded one, not assembled one from a template. I built one, with original mechanics, custom levels, and characters I designed. It took a weekend. I have no engineering background. I used Claude Code, one of the new generation of AI-assisted development tools, and I described what I wanted in plain language while the system generated the underlying code.
I am sharing this not to be glib about what was accomplished, but because the experience genuinely shifted how I think about what is now possible for organizations like the ones I work with. If someone with no technical background can build a working interactive application over the course of two days, the resource constraint that has historically prevented the community development sector from building the digital tools it needs has fundamentally changed.
That change arrived faster than most of us anticipated. And the organizations that find ways to engage with it early will have an advantage that compounds over time.
What Just Happened to the Cost of Building
In early 2025, the linguist and AI researcher Andrej Karpathy coined the term ‘vibe coding’ to describe a new way of building software: describing what you want in natural language, iterating with an AI system that generates and refines the underlying code, and testing the results in real time rather than writing syntax by hand. The phrase was playful, but what it described was genuinely significant. Collins Dictionary named it their Word of the Year for 2025.
The numbers from the startup ecosystem tell part of the story. In Y Combinator’s Spring 2025 cohort, roughly 25% of participating companies reported that 95% or more of their codebase had been generated by AI tools. In 2021, building a basic technology minimum viable product (a functioning first version of a digital product) typically cost between $50,000 and $150,000 and took three to six months. By early 2026, that same scope of work can be completed over a weekend using a combination of AI development tools and a cloud API subscription.
Matt Shumer, a technology founder, published an essay in February 2026 titled ‘Something Big Is Happening’ that articulated this shift in terms practitioners outside the technology sector could understand. Within weeks it had been read more than 80 million times. The breadth of that readership reflects something real: people across many fields are beginning to recognize that the cost and complexity of building digital tools has dropped by an order of magnitude, and that the implications extend well beyond Silicon Valley.
| In 2021, building a basic technology MVP cost between $50,000 and $150,000 and took three to six months. By early 2026, that same scope of work can be completed over a weekend. |
What This Means for Community Development
The first piece in this series described the technology gap in the CDFI and economic development sector, the ways in which organizations built around mission have not kept pace with the digital infrastructure that mission increasingly requires. One of the structural causes of that gap has been resource: building technology was expensive, required specialized talent, and took longer than most grant cycles could accommodate.
That constraint has not disappeared, but it has changed shape significantly. The question is no longer primarily whether the sector can afford to build. It is whether the sector has developed the leadership fluency, the organizational permission structures, and the appetite for experimentation to take advantage of what is now within reach.
Consider what the loan readiness process looks like in most community lending organizations today. A prospective borrower calls or emails to inquire about a loan. A loan officer sends them a checklist of required documents. The borrower, who may be a sole proprietor running their business during the hours the lender’s offices are open, navigates that checklist on their own, often without knowing which documents are most important, which gaps are likely to be dealbreakers, and which issues could be resolved before they submit. Many borrowers drop off at this stage, not because they are not creditworthy, but because the process is opaque and the guidance is not there when they need it.
Now consider what is possible to build today. An AI-assisted loan readiness tool that converses with a borrower in plain language, helps them understand what lenders look for, connects to open banking APIs and accounting platforms like QuickBooks to pull relevant financial data, identifies gaps and suggests ways to address them, and generates the financial package a lender needs, formatted correctly and ready for underwriting review. This is not speculative. Every component required to build this tool exists today. The main thing standing between the sector and that kind of tool is the decision to build it.
I am describing this openly, in a public post, because I think it should exist, and I think the organization or coalition that builds it first will do enormous good. Consider this a provocation and an invitation.
What Organizations Can Do Right Now
The shift in building costs creates an opportunity, but opportunities do not convert themselves. Here is what it looks like when organizations move from awareness to action in this environment.
The most valuable first step is creating protected space for experimentation. That means budget, even a small one, specifically designated for trying things, with leadership’s explicit acknowledgment that not all experiments will work and that uninstructive failure is acceptable. Without that protected space, experimentation gets crowded out by operational demands every time. With it, practitioners have permission to try things that might not work, learn from what they discover, and bring those findings back to inform larger decisions.
The second step is building firsthand fluency at the leadership level. This does not mean executive directors need to become engineers. It means spending a few hours with an AI development tool, building something small, something that does not matter, something just to understand what the process actually feels like. The leaders who have done this tend to ask fundamentally different questions in technology conversations than those who have not. The gap between the two is narrower than it might seem, and the investment required to close it is much smaller than it was even two years ago.
The third step is cross-organizational sharing. One of the genuine structural advantages of the community development sector, compared with private sector competitors, is that organizations working in adjacent markets with shared missions have almost no reason to treat technology as proprietary. A loan readiness tool built by one CDFI could, with modest adaptation, serve ten others. A data integration solution developed by a regional economic development organization could be open-sourced and adapted by counterparts in other cities. The sector has historically underinvested in the shared infrastructure that would make this kind of collaboration systematic. That is beginning to change, and it is worth accelerating.
| One of the genuine structural advantages of the sector is that organizations with shared missions have almost no reason to treat technology as proprietary. A tool built by one CDFI could serve ten others. |
What Funders Can Do
The shift in building costs changes the math for funders too, including how existing commitments should be evaluated and how new ones should be structured.
The most direct opportunity is capacity grants that include a technology mandate. The OFN CDFI Innovation Initiative, which launched in 2025, is an example of what this can look like at scale: patient capital specifically designated for building operational and technology capacity, with evaluation criteria that reward intelligent experimentation rather than just deliverable completion. More of this is needed, and more funders have the resources to provide it than currently do.
Convening and co-investment are also underutilized levers. A foundation that brings ten CDFIs together to define shared technology requirements and co-fund a shared tool gets significantly more impact per dollar than one that funds ten separate technology projects in isolation. The coordination cost is real, but it is far smaller than the waste embedded in ten organizations each trying to solve the same problem independently.
Finally, digital maturity is worth including as a criterion in proposal evaluation, not as a threshold requirement, but as a factor. Organizations that have demonstrated the capacity to steward technology investments well are more likely to deploy new capital effectively. Making that visible in proposal processes creates a productive incentive for organizations to develop that capacity over time.
The MacKenzie Scott funding model, which provides large unrestricted grants on the basis of organizational trust, has produced real transformation for some organizations. It has also, in some cases, provided significant capital to organizations whose leadership and culture were not positioned to deploy it toward capability-building. Unrestricted capital is most powerful when the organization receiving it already has the internal fluency to know what it needs. Building that fluency is the work that happens alongside capital, not downstream of it.
The Window Is Open
Every major shift in technology costs creates a window, a period during which the organizations that move earliest establish structural advantages that compound over time, and during which the gap between early movers and late adopters begins to widen. That window does not stay open indefinitely.
The community development sector has historically entered these windows late, often a decade or more after the cost inflection had already occurred. The fintech wave of the 2010s is one example. The first piece in this series explores those structural reasons in detail. But the structural reasons do not change the outcome for the communities the sector serves when technology that could have reached them earlier does not.
The window is currently open. The tools are more accessible than they have ever been. The evidence base for what works is beginning to mature. The peer networks and funder infrastructure to support experimentation are beginning to develop. The question is whether the sector will engage with this moment with the same seriousness and intentionality it brings to the rest of its work, and what it would take to make that happen at scale.
The final piece in this series looks at the behavioral patterns that have historically determined how the sector responds to moments like this, and at what it would take to make the response more durable than it has been in the past.
