Build vs. Buy: purchasing software in the age of agentic programming

It's no surprise that in the advent of LLM's like Claude, Codex, and open weight models, software has dynamically changed. We've seen the paradigm shift from code complexity being a blocker to most SAAS business ideas buildable from any person, whether they know how to code or not. This assuredly has had an impact on venture capital and ultimately what the "value of software" really is.

By the numbers

• 84% of developers now use or plan to use AI tools in their development process, up from 76% a year earlier; 51% of professional developers use them daily. Trust lags adoption: 46% actively distrust AI output accuracy versus 33% who trust it. [1]
• AI took close to 50% of all global venture funding in 2025, up from 34% in 2024. Foundation-model labs alone raised about $80 billion, 40% of all AI funding. [2]
• Public markets repriced software on the same thesis. After AI agent launches in early 2026, the iShares Expanded Tech-Software Sector ETF fell more than 14%, its worst stretch since 2008. A Jefferies trader dubbed the selloff the "SaaSpocalypse." [3]

I'm going to separate software companies into multiple cohorts:

Cohort 1: Open source software

This is the fundamental building block on all of software. It's how I learned to program many many years ago. It's a collaborative effort with many stakeholders democratizing a shared goal.

Pros:

1. It's free to use (when the licensing permits). This is the biggest one. It's free to build on top of. It's been the backbone of many profitable companies. Some of that functionality is even core to the business.
2. It reduces your cost basis when trying to run a lean business. If a business can build something to solve a problem versus buying it, it can save money.

Cons:

1. You are at the will of the maintainers. You might even have to be one of them. That reduction in cost is shifted to maintenance, which is no small task.
2. Complexity is much higher since the systems are hand rolled and internal. Can cause your business to suffer slower product delivery cycles.

By the numbers

• Harvard Business School researchers put the demand-side value of widely used open source at $8.8 trillion; without it, firms would spend about 3.5 times more on software than they do now. The cost to write that same code once is only about $4.2 billion. [4]
• That value is concentrated: about 5% of contributors account for more than 90% of it. [4]
• Open source appears in 97% of commercial codebases, with about 911 open source components per application on average. [5]
• The maintenance bill is real: 86% of audited codebases contained vulnerable open source and 90% had components more than four years out of date. [5]
• 60% of maintainers describe themselves as unpaid hobbyists. Direct payments from companies make up just 5% of maintainer income, and the share of maintainers under 26 fell from 25% in 2021 to 10% in 2024. [6]

Cohort 2: VC or funding-backed software companies

Speaking from experience, these companies more-or-less follow the same playbook. Inject money, focus on high growth and adoptability in hopes of supernormal returns later. The good ones have competitive advantages in some way that differentiate themselves from their competition, which eventually lead to IPOs or acquisitions. Decisions are usually driven by the investors.

Pros:

1. Usually a first-mover in a market. Innovates in some way on an existing problem.
2. Pricing is usually low because they can operate with negative earnings while growing.

Cons:

1. Focused on growth and not client retention.
2. No guarantee they will succeed. Their cost of operation may exceed their value in the market.
3. Many eventual competitors. AI Risk to have no competitive advantage.

By the numbers

• Low pricing is funded by losses. 48% of equity-backed private SaaS companies operate at a loss, versus 17% of bootstrapped ones; the median equity-backed company spends 101% of its ARR. [12]
• Growth is bought, not retained. Equity-backed companies spend 89% more on sales and 100% more on marketing than bootstrapped peers. [14]
• Most don't pay off. In a study of more than 2,000 companies that raised at least $1 million between 2004 and 2010, HBS lecturer Shikhar Ghosh found about 75% never returned cash to investors; in 30–40% of cases investors lost everything. [7]
• Shutdowns on Carta rose to 966 in 2024 from 769 in 2023, up 25.6%; enterprise SaaS was the largest category, at 32% of closures. [8] Among SimpleClosure wind-downs in the first half of 2026, B2B SaaS was 27.3% of closures, and the median company shut down 2.28 years after its last raise. [18]
• The competition is capitalized. AI companies took $355.9 billion of the $412.7 billion invested in U.S. startups in the first half of 2026. [18] 58% of 2025 AI funding went into megarounds of $500 million or more. [2]

Cohort 3: Public companies

These are your Adobe's, Slack, Figmas. Large bureaucratic entities at the behest of shareholders that expect infinite growth. Change is often slow, and the price is usually high for licensing.

Pros:

1. Stable software. WYSIWYG.

Cons:

1. Enterprise offerings will be expensive.
2. Very little innovation.
3. Choices often don't benefit the consumer.

By the numbers

• Software prices are rising far faster than general prices. Vertice's SaaS Inflation Index put SaaS price growth at 11.4% year over year in January 2025, against 2.7% average inflation in G7 countries. Its preliminary September 2026 reading was 14.3%. [9]
• Investors are now pricing in AI risk to incumbents. On February 3, 2026, an Anthropic legal plug-in for Claude helped trigger a global software selloff; London Stock Exchange Group fell nearly 13% that day. [10] By late March, Adobe was down 32.2% and Salesforce 31.3% year to date, and the iShares Expanded Tech-Software Sector ETF had its worst stretch since 2008. [3]

Cohort 4: Companies that bootstrap themselves

These are companies that trade capital injection for slower growth. These companies focus on profitability over growth. Again, software is a lot easier to build now, so these low-cost, customer-focused software companies will continue to emerge, which adds pressure to the other two cohorts.

Pros:

1. Generally very problem/solution oriented.
2. Focused on profitability and "stickiness".
3. Pricing can be lower than even VC backed companies.

Cons:

1. Can be innovative, but generally solve red ocean problems where competition is higher.
2. Slower product life-cycle with smaller teams.
3. Initially priced out of many marketing and growth tools.

By the numbers

• Bootstrapping is the norm, not the exception: less than 1% of new U.S. firms are funded by venture capitalists. [19]
• The growth gap is small. In SaaS Capital's survey of 1,000+ private B2B SaaS companies, bootstrapped firms grew at a 20% median versus 25% for equity-backed firms. [11]
• The profitability gap is large: 83% of bootstrapped companies are profitable or within two points of breakeven, against 52% of equity-backed companies. [12]
• Stickiness shows in retention. Bootstrapped companies with $3M–$20M in ARR report median net revenue retention of 103%. [13]
• The marketing gap is real: equity-backed companies spend 89% more on sales and 100% more on marketing than bootstrapped peers. [14]

Who to pay and why to pay and what you get

The disruption of the advent of AI is the perfect time to step back and evaluate. As the market corrects itself, we will see much more value in the teams you're actively supporting with more hyper-tailored SAAS products.

By the numbers

• The people are the expensive part. The median U.S. wage for software developers, QA analysts and testers was $134,040 in May 2025, against $50,980 for all occupations. [15]
• Building is only the down payment. Maintenance averages about 60% of a system's lifecycle cost (40–80% depending on the system), and about 60% of that goes to user-requested changes rather than bug fixes. [16]
• AI does not yet erase that labor. In a 2025 randomized trial, 16 experienced open source developers took 19% longer on 246 real tasks when allowed AI tools, while believing they had been 20% faster. They accepted fewer than 44% of AI-generated code. [17]

A worked example: three ways to buy the same category

Take one narrow category: desktop apps that run a fleet of coding agents (Claude Code, Codex, Cursor, OpenCode) in parallel git worktrees on your own machine, using the model subscriptions you already pay for. Three products do roughly the same job. They come from three different cohorts, and that is the whole point.

Conductor is the Cohort 2 playbook exactly as described above. Free to the user, funded by $22 million of outside money, and first-mover in the category when it launched in summer 2025. [21] Seven months after the Series A, the team joined Meta Superintelligence Labs. The product was never the business; the team and the interface were. Users of Conductor Cloud got 16 days' notice before their data is deleted. [23] This is not a knock on the founders. It is what the playbook is for. The acquisition is the success case.

Orca is the open source flavor of Cohort 2. It is free, well built, and MIT-licensed, which means you can fork it if the company changes direction. [22] But it is not the company's product; Stably's revenue is an AI testing platform, and Orca is the funnel. The Cohort 1 cons apply: you are at the will of the maintainers, and the maintainers' incentive is a different product. If Stably is acquired, pivots, or runs out of runway, you own a codebase, not a roadmap.

Scape is Cohort 4. There is no investor to return cash to and no acquirer to satisfy. The $99 a year is the entire business model, so the only way Scape survives is if people who run agent fleets keep finding it worth $99. [20] That is the alignment you are paying for. The feature set reflects it: Argus orchestration, playbooks and automations you can switch models under without rebuilding, adversarial reviews between models, and integrations into Jira, Linear, GitHub, Slack and Datadog, plus support from the people who wrote it. [20] Scape is paid software because the alternative is being the product for someone else's exit.

The lesson is not that free is bad. Conductor was good software. It is that "free" in Cohort 2 is a price set by the investor's timeline, not yours, and the bill comes due the day the timeline ends. In the SaaS Capital data, 48% of equity-backed private SaaS companies run at a loss versus 17% of bootstrapped ones. [12] When you choose the tool that runs your agents, your automations, and your playbooks, you are choosing which of those two numbers you want to depend on.

Works cited

1. Stack Overflow. 2025 Developer Survey: AI. July 2025.
2. Crunchbase News. 6 Charts That Show The Big AI Funding Trends Of 2025. December 2025.
3. The Motley Fool. Software Stocks Entered a Bear Market. March 27, 2026.
4. Hoffmann, M., Nagle, F., and Zhou, Y. The Value of Open Source Software. Harvard Business School Working Paper 24-038, 2024. Summary: HBS Working Knowledge.
5. Black Duck. 2025 Open Source Security and Risk Analysis (OSSRA) report, as reported by SC Media. February 2025.
6. Tidelift. 2024 State of the Open Source Maintainer Report, as reported by The Register. September 18, 2024.
7. Ghosh, S. (Harvard Business School), as reported by GeekWire. September 2012.
8. TechCrunch. 2025 will likely be another brutal year of failed startups, data suggests. January 26, 2025.
9. Vertice. Key SaaS inflation stats (2025) and SaaS inflation rate tracker (accessed November 2026).
10. Reuters, via The Star. Global software stocks hit by Anthropic wake-up call on AI disruption. February 4, 2026.
11. SaaS Capital. 2026 Private B2B SaaS Company Growth Rate Benchmarks. 2026.
12. SaaS Capital. 2026 Spending Benchmarks for Private B2B SaaS Companies. June 10, 2026.
13. SaaS Capital. 2026 Benchmarking Metrics for Bootstrapped SaaS Companies. April 24, 2026.
14. SaaS Capital. 2025 B2B Benchmarking: Private SaaS Company Growth Rates. June 2025.
15. U.S. Bureau of Labor Statistics. Occupational Outlook Handbook: Computer and Information Technology Occupations. May 2025 wage data.
16. Wood, D. The 60/60 Rule. In 97 Things Every Project Manager Should Know, O'Reilly, 2009.
17. METR. Measuring the Impact of Early-2025 AI on Experienced Open-Source Developer Productivity, as reported by ITPro. July 2025.
18. SimpleClosure. State of Startup Shutdowns: H1 2026. 2026.
19. Kauffman Foundation. Entrepreneurship Policy Digest, as announced on PRWeb. June 2015.
20. Scape. scape.work (accessed November 4, 2026).
21. TechCrunch. Meta acquires AI coding startup Conductor. November 4, 2026.
22. Stably AI. Orca repository and README (MIT license); company profile via Dealroom.
23. Runtime Wire. Meta acquires Conductor and sets a November 20th cloud shutdown. November 2026.