The AI product cycle is compressing faster than most leaders can process. The pace of AI progress has moved from fast to disorienting. That sounds dramatic until you look at the last few weeks.
Claude Design launched. Google pushed Stitch forward and open-sourced DESIGN.md. Andrej Karpathy, a co-founder of OpenAI, said he has barely typed a line of code since December. Garry Tan, CEO of Y Combinator, pointed to solo builders shipping with the leverage of entire teams. DeepSeek V4 arrived with a Huawei-aligned infrastructure story. Anthropic’s Claude Mythos Preview and Project Glasswing put frontier AI into the security conversation for critical software. Reuters reported that SpaceX has an option structure around Cursor that could value the company at roughly $60 billion. (Anthropic)
That is a ridiculous amount of movement for one slice of the market.
And it all points in the same direction.
The product-making cycle is collapsing.
By collapse, I do not mean failure. I mean compression. The old sequence of strategy, research, design, engineering, QA, security, and iteration is becoming one tighter loop. The distance between idea, artifact, prototype, production, and feedback is shrinking.
That is the story.
Last August, I wrote Vibe Everything: The Singularity That Will Eat Product. My thesis was that strategy, design, engineering, validation, and iteration would converge into one AI-assisted operating loop.
I believed that was coming.
I did not think it would arrive this fast.
A few days ago, I published Claude Design and AI Native Product Creation on the Bootcamp UX blog. I wrote about how Claude Design starts to democratize the first credible product artifact: the screen, prototype, flow, deck, or visual expression that turns an abstract idea into something people can react to.
Then Google moved.
Google Saw The Same Opening
A few days after Anthropic launched Claude Design on April 17, Google pushed the same conversation forward from another angle.
Google had already introduced Stitch as an AI-powered design tool for generating UI from prompts and visual inputs. The bigger move came on April 21, when Google open-sourced the draft specification for DESIGN.md, a format meant to carry design rules across tools and platforms.
That is the signal.
AI design needs more than prompts. It needs shared rules, persistent context, reusable systems, accessibility constraints, and machine-readable design intent.
This should matter to every tech executive who has funded a design system and wondered why it still feels disconnected from production.
In an AI-native product environment, the design system becomes an operating layer for product creation. If agents are going to design screens, write code, generate variants, create prototypes, and hand work across the stack, they need the rules of the business embedded in the system.
Brand. Accessibility. Interaction patterns. Component logic. Product principles. Technical constraints.
The value goes beyond faster mockups. The real value is a product organization that can convert intent into artifacts, artifacts into prototypes, and prototypes into production work with less translation loss.
Less translation loss means faster decisions, fewer rework cycles, lower delivery cost, and better odds that teams spend money on work that actually reaches customers.
This is one reason the AI product cycle now depends on machine-readable design intent, not just better interfaces.
The Developer Workflow Already Changed
The same shift is happening in software development.
In March 2026, Andrej Karpathy, a co-founder of OpenAI, said something on the No Priors podcast that should be taped to every CTO’s monitor:
“I don't think I've typed like a line of code probably since December, basically, which is an extremely large change.”
That quote matters because Karpathy was describing a workflow change. He said his default software-building workflow had become completely different since December.
The developer is becoming less of a typist and more of an orchestrator. The best engineers are moving up the abstraction stack. They are defining intent, setting constraints, reviewing output, managing agents, and deciding what deserves to survive contact with production.
This changes the unit economics of software creation.
A strong builder with design taste, technical judgment, and the right agent setup can now produce output that previously required a much larger team. The shape of the team changes. Smaller groups can cover more surface area. Senior judgment becomes more valuable. Weak judgment becomes more expensive.
Garry Tan made this point bluntly in the README for gstack, his open-source Claude Code setup. Tan is a Canadian-American venture capitalist, executive, and CEO of Y Combinator. He quotes Karpathy, then asks: “How does one person ship like a team of twenty?” He points to Peter Steinberger building OpenClaw, which Tan describes as a 247K-star GitHub project built essentially solo with AI agents. His conclusion is the line tech executives should pay attention to: “A single builder with the right tooling can move faster than a traditional team.” (GitHub)
This is a significant milestone because it reframes productivity from team size to system leverage.
For decades, the default answer to more product ambition was more headcount. Now the better question is whether the team has the right tooling, context, architecture, test coverage, design rules, and decision discipline to let a small number of strong people move with massive leverage.
Treat the GitHub star count as a moving number. The strategic signal is still useful.
The old assumption was that scale required headcount. The new assumption is that scale increasingly depends on tooling, architecture, context, systems, and judgment.
Headcount still matters. Coordination drag matters more because every additional handoff creates delay, translation loss, meeting load, and accountability blur. AI makes creation faster, which means organizational drag becomes more visible and more expensive.
Vibe Coding Was The Doorway. Agentic Development Is The Evolution.
The first phase of this shift was vibe coding.
Describe what you want, let the model generate the code, keep nudging until something works. That is vibe coding in a nutshell. Tools like Cursor made that behavior visible to a much larger market by bringing AI coding into the developer workflow.
It collapsed the distance between idea and artifact. It let more people build. It gave founders, designers, product managers, and engineers a faster way to explore.
It also exposed the danger immediately.
Software that looks finished can still be full of security issues and bugs, brittle, expensive to maintain, poorly architected, inaccessible, and impossible to scale. A prototype can create false confidence. A demo can seduce leadership into thinking the hard work is done.
The AI product cycle gets faster when tools can understand the repo, run tests, create pull requests, and keep humans focused on judgment.
Agentic development is the evolution.
It adds structure: repositories, tests, permissions, plans, code review, tool access, observability, security checks, context files, design rules, documentation, automated QA, and deployment workflows.
That is where Claude Code, Codex, Stitch, DESIGN.md, gstack, Cursor, and other agent harnesses start to form a pattern. Claude Code lives close to the codebase. Codex works as a cloud-based software engineering agent that can work on tasks in parallel, write features, fix bugs, answer codebase questions, and propose pull requests. (OpenAI)
Different products. Same direction.
The product-making loop is collapsing.
Mythos Shows Why Security Has To Move Into The Loop
This is where Anthropic’s Claude Mythos Preview matters.
Mythos Preview is an unreleased frontier model that Anthropic says can surpass all but the most skilled humans at finding and exploiting software vulnerabilities. Anthropic says it found thousands of high-severity vulnerabilities, including vulnerabilities in every major operating system and every major web browser. (Anthropic)
That is not a normal product announcement.
That is a warning flare.
Anthropic’s own red-team writeup says Mythos Preview found a 27-year-old OpenBSD vulnerability, a 16-year-old FFmpeg vulnerability, and chained together vulnerabilities in the Linux kernel. It also wrote a browser exploit that escaped both renderer and OS sandboxes, and autonomously wrote a remote code execution exploit against FreeBSD’s NFS server that granted full root access to unauthenticated users. (Anthropic)
Pause on that.
We are talking about operating systems, media libraries, kernels, browser sandboxes, and network software. This is the infrastructure modern companies, cloud platforms, products, and the internet itself rely on.
Anthropic says Project Glasswing gives selected partners access to Mythos Preview so they can find and fix vulnerabilities in foundational systems. The partner list includes AWS, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks. Anthropic also says it extended access to more than 40 additional organizations that build or maintain critical software infrastructure. (Anthropic)
The business point is straightforward.
AI will not only accelerate product creation. It will accelerate the discovery of weakness in the systems products depend on.
That changes the operating model for every serious technology organization.
Security cannot live at the end of the process. It has to sit inside the product loop. The same system that helps generate code needs to help inspect code, test code, automate QA, validate dependencies, monitor runtime behavior, improve observability, and surface risk before production.
For tech executives, AI adoption needs operating rules.
Define where agents can operate.
Define what they can access.
Define which actions require human approval.
Define how generated code gets tested.
Define how automated QA gets trusted.
Define how observability data gets used to detect regressions.
Define how dependencies get scanned.
Define how security issues get triaged.
Define who owns the final call.
AI is a force multiplier. It multiplies your strengths and your weaknesses.
DeepSeek Makes This Bigger Than Product Teams
Today is April 24, 2026.
Reuters reported this morning that DeepSeek launched a preview of its new V4 model, adapted to run on Huawei chip technology. Reuters also reported that DeepSeek says V4 is particularly suited to AI agent work, which is the layer now driving software creation, research, testing, and automation. (Reuters)
That matters beyond benchmarks.
DeepSeek V4 is geopolitical because model adoption is platform power: whoever owns the default tools for building software gains influence over infrastructure, standards, cost expectations, and the behavior of the next generation of products.
If a large share of the world’s developers, startups, enterprises, and governments build on Chinese models, China gains influence over more than AI adoption. It gains influence over the operating layer of modern software.
That does not require conspiracy thinking.
That is how platforms work.
The dominant platform shapes the market around it.
Reuters also reported that DeepSeek’s close collaboration with Huawei contrasts with its past reliance on Nvidia chips, and that Washington’s chip restrictions have accelerated Beijing’s push for tech self-sufficiency. Huawei also said its Ascend supernode based on Ascend 950 AI chips would support DeepSeek’s V4 versions. (Reuters)
This is the bigger tech executive lesson: the AI race is now full-stack and geopolitical.
Models, chips, developer tools, agents, design systems, cloud platforms, security workflows, national strategy, and capital markets are starting to move together.
Cursor Shows Where The Money Is Going
That is why the reported SpaceX and Cursor structure matters.
Cursor is an AI coding environment built around codebase understanding, agents, code review, and developer workflow. Cursor describes itself as “the best way to code with AI,” and its site says agents can turn ideas into code while developers focus on decisions. (Cursor)
Reuters reported today that SpaceX has the option to buy Cursor for about $60 billion, or walk away and pay roughly $10 billion for a collaboration. Reuters frames Cursor as an AI code-generation startup and says the structure allows SpaceX to delay a decision until after its IPO. (Reuters)
Take the structure seriously. Reuters describes it as an option structure.
The signal is still obvious.
AI coding tools are becoming critical infrastructure. Companies with ambition, capital, compute, and distribution are moving toward the software creation layer because that layer controls speed.
And speed compounds.
The Business Case: AI Product Cycle Needs Velocity With Control
The tech executive question is simple:
Where can AI compress cycle time, improve quality, and protect margin without increasing enterprise risk?
Use AI to reduce the cost of learning. Use it to generate more options earlier. Use it to turn ideas into prototypes. Use it to inspect code. Use it to automate QA. Use it to improve documentation. Use it to find defects. Use it to strengthen security. Use it to improve observability. Use it to optimize the foundational infrastructure layer: compute usage, deployment pipelines, dependency management, incident detection, reliability, and system performance.
Then put real controls around the work.
The revenue impact comes from faster product learning, faster shipping, better personalization, better conversion, better retention, and more experiments that reach customers.
The profitability impact comes from lower rework, lower coordination cost, faster QA cycles, better infrastructure efficiency, fewer production failures, and less time wasted turning one team’s intent into another team’s backlog.
The value of AI in product creation will show up in business outcomes: adoption, retention, conversion, revenue impact, cost reduction, cycle time, quality, and risk reduction.
The leaders who win will connect AI adoption to those outcomes.
The leaders who lose will confuse activity and tooling with progress.
The AI Product Lifecycle Is Becoming One Loop
We are actively defining a new product operating model where strategy, design, engineering, testing, security, and iteration converge into one AI-assisted execution loop.
Claude Design shows where visual product creation is heading. Google’s DESIGN.md shows why design systems need to become machine-readable. Karpathy’s workflow shows how software development is moving from typing to orchestration. Garry Tan’s gstack points to a world where one strong builder can operate with the leverage of a much larger team.
Claude Code, Codex, Cursor, gstack, and other agent harnesses show how software creation is moving from individual prompting to managed execution systems.
Mythos and Project Glasswing show why security has to move into the same loop.
DeepSeek V4 shows that the model and hardware race is accelerating. It also shows that AI leadership is now tied to national strategy, infrastructure independence, and global platform influence.
The reported SpaceX and Cursor structure shows how strategically valuable AI coding tools have become.
This is a small slice of what has happened in the last few weeks.
The direction is clear.
Ideas become artifacts faster. Artifacts become prototypes faster. Prototypes become code faster. Code gets reviewed, tested, secured, observed, optimized, and shipped through increasingly agentic systems.
This is the most exciting moment in product creation I have seen in years.
It is also the moment where leadership discipline matters most.
The tools are accelerating.
So here is the uncomfortable question for every tech executive: are you redesigning the operating model around this new speed, or are you just buying tools and hoping the org magically gets faster?
What This Means For Sports Tech
For the sports executives in my network, this is where the rubber meets the road.
Sports is the perfect stress test for AI-native product creation: live events, emotional users, fragmented data, rights constraints, sponsorship demands, betting adjacency, media workflows, commerce, personalization, and zero patience for broken experiences when the game is on.
The product cycle collapsing will hit sports harder than most industries because the window to create value is brutally short. A fan moment appears, peaks, and disappears. The organizations that can turn insight into experience fastest will have the advantage.
This has major implications for fan engagement, media operations, ticketing, sponsorship activation, personalization, loyalty, commerce, venue experience, athlete content, and real-time storytelling.
I’m putting together a deeper piece on what this means for sports technology specifically.
If you lead, invest in, build, or operate in tech, stay tuned here (or on linked in and medium)
The next piece is about where this gets practical: who gains advantage, who gets exposed, and how sports organizations should build when AI compresses the distance between moment, product, and revenue.
Originally published on Linkedin on April 24th, 2006
Keep Reading
If this article was useful, these pieces build on the same argument from different angles:
Vibe Everything: The Singularity That Will Eat Product
The earlier thesis behind this piece: AI is compressing strategy, design, and build into one faster product loop.
Claude Design Means Design Just Lost Its Monopoly on the First Draft
A closer look at how AI-native design tools are changing who gets to create the first credible product artifact.
AI Broke the Interface. Now What?
Why AI is pushing UX beyond static screens and into conversations, agents, trust, and real-time interaction.
Why Designers Keep Failing at the Finish Line
Why design work dies when it ignores the software development lifecycle and the realities of production.
Design is About Business Outcomes
A direct argument for treating design as a business lever tied to adoption, retention, conversion, and revenue.
Why Being Great at One Thing Isn’t Enough Anymore
Why modern product leaders need fluency across design, engineering, strategy, and business.
The pace is only getting faster. I’ll keep writing about where AI, product strategy, design, engineering, and digital business are heading next.
Start with Vibe Everything if you want the broader thesis behind this article.
