Thursday, February 19, 2026
AI Capital Flywheels and the SaaS-pocalypse: Why Coding is Solved but Distribution is the New Moat
The Big Picture
- The $1B ASIC Threshold — Martin Casado explains that when a single AI training run hits $1 billion, a 20% efficiency gain ($200M) makes custom silicon development economically rational.
- Coding is Largely Solved — Boris Cherny reports that 4% of all public GitHub commits are now authored by Claude Code, shifting the professional role from 'Engineer' to 'Builder.'
- The Code Slop Crisis — Sarah Guo warns that 'vibe coding' creates fragile codebases that no human deeply understands, necessitating new tools for engineering risk management.
- The Gray Rock Protocol — Rob Dial details a method for neutralizing toxic manipulators by becoming intentionally uninteresting and non-responsive to emotional triggers.
- CNS Recovery Metric — Jeff Cavaliere introduces the 10% grip strength drop-off rule as a definitive objective signal to skip training and avoid systemic overtraining.
- Virtual Safety Labs — Károly Zsolnai-Fehér showcases chemically rigorous fire simulations that allow for predictive 'what-if' safety testing in real-time environments.
The Deeper Picture
The technology landscape is currently defined by a fundamental shift in the unit of work, moving from manual craft to high-level orchestration. In Inside AI’s $10B+ Capital Flywheel — Martin Casado & Sarah Wang of a16z, we see the emergence of a 'Capital Flywheel' where dollars are directly converted into intelligence gains via scaling laws. This massive influx of capital is enabling a reality where, as argued in Head of Claude Code: What happens after coding is solved | Boris Cherny, software engineering is entering its 'Printing Press' moment. Coding is becoming a universal capability rather than a guild skill, evidenced by the fact that AI agents already author 4% of public GitHub commits.
However, this rapid democratization of code production introduces a secondary crisis: the 'Code Slop' problem. As discussed in The AI Code Slop: Risk or Opportunity, the ability to 'vibe code' vast amounts of software creates fragile, black-box codebases that no human architect fully understands. This mirrors the shift in physical sciences seen in The Most Realistic Fire Simulation Ever, where we are moving from 'pretty pixels' to chemically rigorous 'virtual safety labs.' In both domains, the value is shifting from the act of creation to the act of verification and risk management.
This high-velocity environment demands peak cognitive and physical performance. Just as Jeff Cavaliere suggests in Essentials: Optimize Your Exercise Program with Science-Based Tools that we must monitor the central nervous system via grip strength to avoid overtraining, Rob Dial argues in This Is How Smart People Handle Toxic People that we must protect our mental bandwidth through the 'Gray Rock Method.' The common thread across these disciplines is the transition to a 'rule-based' rather than 'suggestion-based' approach to managing complex systems, whether they are AI models, toxic relationships, or physical recovery.
Where Videos Converge
The Shift from Engineer to Builder
Head of Claude Code: What happens after coding is solved · The AI Code Slop: Risk or Opportunity · Inside AI’s $10B+ Capital Flywheel
There is a clear consensus that the manual act of writing code is being commoditized. The value is migrating toward product vision, distribution, and the orchestration of agentic workflows rather than syntax mastery.
Unprecedented Revenue Velocity
Inside AI’s $10B+ Capital Flywheel · The AI Code Slop: Risk or Opportunity
Both videos highlight that AI companies are reaching $10B in revenue in roughly one year, a milestone that took traditional SaaS giants like Salesforce or Adobe nearly a decade. This compresses the 10-year displacement cycle into 1-2 years.
Key Tensions
The Future of the Software Engineer Title
Boris Cherny
The title 'Software Engineer' is largely solved and will be replaced by 'Builder' as coding becomes a universal capability.
Sarah Guo
While code generation is easy, the structural advantages of enterprise software (security, change management) mean the engineering role shifts to risk management rather than disappearing.
Resolution: The role likely bifurcates: 'Builders' handle rapid prototyping and product-market fit, while 'Risk Managers' or 'Architects' handle the integration and safety of the resulting 'code slop' in enterprise environments.
Video Breakdowns
6 videos analyzed
Inside AI’s $10B+ Capital Flywheel — Martin Casado & Sarah Wang of a16z
Latent Space · Martin Casado, Sarah Wang · 55 min
Watch on YouTube →The AI industry has entered a 'Capital Flywheel' where dollars are directly converted into capability gains via scaling laws. This has led to $1B training runs that make custom silicon economically viable and a market dynamic where foundation models may consume their own app ecosystems.
Logical Flow
- The Venture-Growth Convergence
- Scaling Laws as Capital Levers
- The ASIC Economic Threshold
- The Expanding Star Theory
- Boring Enterprise Software Opportunity
Key Quotes
"There's no dark GPUs... if someone invests in a company that will actually use the GPUs, on the other side of it is the actual customer."
"If Anthropic can raise three times more every subsequent round, they probably can raise more money than the entire app ecosystem that's built on top of it."
Key Statistics
$1 Billion training run cost
$10 Million talent compensation packages
Weeks to reach tens of millions in ARR
Contrarian Corner
From: Head of Claude Code: What happens after coding is solved | Boris Cherny
The Insight
Underfund your teams to force them to 'Claudify' and automate their own workflows.
Why Counterintuitive
Traditional management suggests that more resources lead to faster output. Cherny argues that scarcity forces engineers to use AI agents as primary executors rather than just assistants.
So What
If you are leading a technical team, resist the urge to hire more heads. Instead, provide unlimited compute/tokens and set goals that are impossible to achieve without agentic automation.
Action Items
Implement a daily grip strength recovery check.
Grip strength is a direct window into central nervous system fatigue.
First step: Use a bathroom scale or dynamometer to establish a baseline. If output drops by 10%, take a rest day.
Deploy the 'Gray Rock Method' with toxic stakeholders.
Neutralizing emotional engagement protects cognitive bandwidth for high-value work.
First step: Identify one 'energy vampire' and commit to giving only short, neutral responses (e.g., 'Okay', 'Yeah') for one week.
Preschedule 'Exit Strategy' board meetings.
In high-velocity AI markets, displacement cycles are compressed, making logical exit planning essential.
First step: Add a recurring agenda item to your board meetings once or twice a year to discuss potential exits while the conversation is non-emotional.
Audit your product for 'Latent Demand' abuse.
Users hacking your tool for unintended purposes is the strongest signal for your next product.
First step: Review your telemetry for 'creative abuse' patterns—where users are using your tool to solve problems it wasn't designed for.
Final Thought
The common thread across today's intelligence is the collapse of traditional bottlenecks—whether in code production, capital deployment, or physical simulation. As the cost of 'doing' approaches zero, the value of 'directing' and 'protecting' becomes paramount. Success in this era requires the builder's mindset to orchestrate agents, the investor's logic to navigate hyper-velocity markets, and the athlete's discipline to protect the underlying cognitive and physical assets.