The Barbell of Software Value
Introduction
Over the past few months, I have been thinking a lot about where software value is moving in the age of AI. In April 2025, Guyren G Howe published a blog called Programming in the Age of Abundance [1]. His argument was that as programming becomes cheaper, we will see software explode across every industry and workflow. He believes abundance means more software everywhere, not less.
I had been working on my own framework, which I called the Internal Software Leverage Theory. My idea was that selling software itself is a weak business model in the age of AI because any feature can be copied quickly. The real moat comes from building proprietary tools that never get sold, tools that only your company uses to operate faster and smarter. I gave the example of a real estate company building its own underwriting and acquisition software. Competitors can never access it, but it quietly compounds advantage.
At first, these two ideas seemed opposed. Howe argues that abundance means expansion, while I argued that selling software is mostly dead. After researching further and looking at real market examples, I think both contain part of the truth. The final framework I have landed on is what I call the Barbell of Software Value.
Glossary
To make sure the discussion is clear, here are a few quick definitions:
- Horizontal SaaS: Generic tools sold across many industries (e.g. Slack, Zoom, Dropbox). Easy to adopt, but also easy to replace.
- Vertical SaaS: Tools built for one specific industry or workflow (e.g. Bild AI for construction). Narrower audience but harder to dislodge once embedded.
- Platform / Primitives: Foundational layers like clouds, models, and developer platforms that others build on. Think AWS, OpenAI, or GitHub.
- Private Internal Leverage: Software built only for your own company, not sold to others, often powered by proprietary data. Examples include JPMorgan’s AI suite or Walmart’s internal agents.
With that in mind, let’s look at the signals from today’s market.
Market Signals
The evidence is already clear in 2025.
Cursor, the AI IDE, recently raised roughly nine hundred million dollars at a nine point nine billion valuation. Some dismiss it as a wrapper since it does not host its own models. Cursor thrives because it goes deeper than a simple UI. It handles repository-scale context, safe multi-file diffs, enterprise security, and cost controls. Competitors like Microsoft/GitHub and Amazon/AWS have copied features through the integration of GitHub Co-pilot and the introduction of the Kira IDE, but the system-level execution, reliability, and developer adoption keep Cursor alive. This is a good example of the platform or vertical end of the barbell [2] [3].
Bild AI, a YC-backed startup, applies vision and language models to construction blueprints. It lets firms bid on more projects with less error. This is vertical SaaS. It is not a platform, nor a private internal system, but it survives because it focuses on a painful domain workflow where ROI is undeniable. Big tech is unlikely to chase something this specific, which gives Bild space to grow [4].
Then there are the giants. Walmart is consolidating its internal agents into four super agents for customers, employees, engineers, and suppliers. JPMorgan reports that its internal LLM Suite is used by about two hundred thousand employees. These are not products to be sold on the market. They are private internal leverage that compounds efficiency inside large firms [5] [6].
At the same time, enterprise surveys show AI adoption rising fast. A McKinsey study reports that seventy-eight percent of companies now use AI in at least one function. Bain reports that ninety-five percent of United States companies use generative AI and production use cases doubled year over year. In parallel, CFOs are cutting SaaS sprawl. Companies want fewer vendors and higher ROI. Generic tools with thin differentiation are the first to be replaced [7] [8].
Lessons from the 2010s
To understand why the middle of the barbell is eroding, it helps to look back at the 2010s. This was the golden era of horizontal SaaS. Companies like Slack and Zoom exploded by selling simple, general-purpose tools that anyone in any industry could adopt. Their distribution advantage came from ease of use, freemium models, and viral adoption inside teams [9].
But fast forward to today, and the landscape is different. Microsoft Teams now bundles chat, meetings, and file sharing directly into Office 365. Salesforce, once the poster child of SaaS, has entrenched itself by moving deeper into verticals and expanding into a full platform ecosystem [10]. The old playbook of “build a horizontal tool and scale” is much harder to repeat because distribution is dominated by incumbents and AI makes feature replication trivial.
This is why the barbell is forming. The middle ground that once fueled SaaS unicorns is shrinking, while the poles (platforms and private leverage) are where durable value now concentrates.
The Barbell of Software Value
Here is the theory as simply as possible. Software value is concentrating at two poles. On one end, you have platforms and primitives like clouds, models, and orchestration layers. These win because everyone builds on them, and big tech dominates much of this space through distribution and bundling. On the other end, you have private internal systems that companies build for themselves and never sell. These are moats because they are wired directly into operations, powered by proprietary data, and impossible for competitors to access.
What erodes is the middle. Generic horizontal SaaS is squeezed by AI-driven feature parity and CFO vendor consolidation. The only SaaS that survives here is highly vertical and ROI-proven, with deep integrations, compliance, or outcomes so strong that switching is painful. If your product cannot clear that bar, it should not be sold. It should be kept private as leverage.
Closing Thoughts
I started with my own theory about internal leverage. Then I read Howe’s blog and forced myself to challenge it. Looking at Cursor, Bild AI, and the moves by Walmart and JPMorgan, the pattern is clear. Howe is right that abundance means more software everywhere. My earlier view was right that selling software gets weaker as a moat. The reconciliation is the barbell. Platforms and primitives on one side, private internal systems on the other, and a middle that only survives if it is extremely vertical and valuable.
For developers and founders, the lesson is straightforward. Pick an end of the barbell. Build at the platform layer or build internal leverage. If you want to sell SaaS, make sure it is vertical, deeply integrated, and ROI-proven. Everything else will get commoditized.
Citations
[1] Guyren G Howe, Programming in the Age of Abundance, Apr 16, 2025.
[2] TechCrunch, Cursor’s Anysphere nabs 9.9B valuation, Jun 5, 2025.
[3] Yahoo Finance, Cursor’s Anysphere nabs 9.9B valuation, Jun 5, 2025.
[4] Y Combinator, Bild AI company page.
[5] CIO Dive, Walmart expands AI leadership and introduces four super agents, Jul 24, 2025.
[6] CIO.com, JPMorgan builds AI foundation on AWS and reports 200,000 employees using LLM Suite, Dec 4, 2024.
[7] McKinsey, The State of AI 2025, PDF, Mar 5, 2025.
[8] Bain & Company, Survey: Generative AI’s uptake is unprecedented despite roadblocks, May 7, 2025.
[9] The Verge, How Slack changed workplace communication, Aug 2019.
[10] Forbes, Salesforce: Still The King Of SaaS, Apr 2023.