{"id":"2072713807462281510","url":"https://x.com/AshwinKadaru/status/2072713807462281510","text":"","author":{"name":"Ashwin","username":"AshwinKadaru","avatarUrl":"https://pbs.twimg.com/profile_images/1664309735636582409/ErVnv8r2_200x200.jpg"},"createdAt":"Thu Jul 02 16:07:35 +0000 2026","engagement":{"replies":4,"retweets":5,"likes":23,"views":42149},"article":{"title":"The Cambrian Explosion","previewText":"There is an explosion happening right now. Inside your company, and it's only getting louder.\n540 million years ago, life on earth went from a handful of simple organisms to nearly every kind of","coverImageUrl":"https://pbs.twimg.com/media/HMO5WrhWsAAcZ2L.jpg","content":"There is an explosion happening right now. Inside your company, and it's only getting louder.\n\n540 million years ago, life on earth went from a handful of simple organisms to nearly every kind of animal that's alive today. Fish, insects, shellfish, the ancestors of everything with a spine. Almost all of it showed up in one short window. In the fossil record it looks instant. Biologists call it The Cambrian Explosion.\n\nIt's happening again. This time with software.\n\nAnd the people setting it off have never written a line of code.\n\nThis is about your job. Whatever your job is. Give me 5 minutes.\n\n## We are at the start, not the middle\n\nKarpathy coined \"vibe coding\" 17 months ago. Collins Dictionary made it word of the year. Claude Code went from zero to about $2.5B in annual sales in roughly 9 months. The fastest any software has ever grown.\n\nSo you'd think the wave already peaked, but those numbers hide something: that was just the engineers and early adopters.\n\nEnterprises are only NOW giving non-technical employees access to tools like Claude Code and Cowork. Partly because the security story wasn't there before.\n\nAnd the moment non-technical people get access, something strange happens.\n\nThey build.\n\nApps to replace the spreadsheets they hate, dashboards shaped around exactly how their team works. Most of it is broken in a dozen ways like no databases, everything crammed into one giant file, or no real development process or version control...\n\nBUT their teams get real value from it anyway. That's not a cautionary tale. That's a fossil record forming.\n\nThe actual Cambrian explosion took millions of years. This one is in year one. Want the simplest proof we're early? Everyone has used AI by now. Almost no one BUILDS with it. That gap is the explosion, and every enterprise rollout of Claude or ChatGPT makes it louder.\n\nHow many people you know has their own app for how they work? We're looking at the start of the curve.\n\n## The age of custom built everything\n\nHistorically, nobody built an app for 3 people. The economics were absurd... You bought SaaS and bent your workflow around someone else's product.\n\nThat constraint just died. Most people haven't noticed what it makes possible yet.\n\nWhen building costs almost nothing, an app for 3 people makes sense. An app for ONE person makes sense. I write custom scripts for one-off tasks because it's cheaper than not doing it. When I'm learning something new, I have Claude build me a mini website, I learn what I need, and I throw it away.\n\nMost of these apps are supposed to die. That's the model.. In the real Cambrian explosion, dozens of radically different animals appeared and most went extinct. The few that survived defined animal life for the next half billion years.\n\nSame math here. 95 people build apps that go nowhere as businesses. 5 build things that change how their team wins deals or ships work. You take that trade every time. The 95 are the cost of finding the 5.\n\nThe obvious pushback: won't ChatGPT and Claude just become the everything app? Maybe. But even then, people will still spin up their own throwaway tools for a few hours and move on. The explosion is the habit of building, not the apps left behind.\n\nAnd none of this comes from SaaS. SaaS is built for the average of a thousand companies. Your people can now build for the specific way YOUR business works.\n\n## The progression ladder\n\nLet me tell you what happens next at your company.\n\n![](https://pbs.twimg.com/media/HMO6-gOWUAApSuS.jpg)\n\nEvery client conversation starts the same way now: \"We rolled out Claude/ChatGPT. What's the ROI? Our spend is climbing.\"\n\nBoards are reading the stat that 95% of company AI projects show no impact on the bottom line. Sounds damning. That's the extinction rate. The question is whether you're set up to find your 5.\n\nThe usage numbers look bad everywhere too. 70% of the Fortune 500 pays for Microsoft's AI assistant. Maybe a third of those paid licenses actually get used in a given week.\n\nBut usage was always the wrong thing to measure. It only tells you whether people are chatting with the AI. The value shows up when people BUILD with it.\n\nThe answer I keep giving clients is two levers, pulled together:\n\n1. Augment the human: Give people a safe way to build, not just chat. Template repos with safety rails and automatic checks built in, trainings, a clear path from idea to finished app. Requests get triaged automatically. Roughly 90% get approved on the spot, and IT only has to look at the 10% that actually needs a human judgment call\n\n1. Augment the AI: That one guarded door connecting Claude to your internal systems. Logins, activity logs, usage limits, and permissions all handled in one place. Now the AI's answers come back with your own company data in them, and the tools your people build can reach something real\n\nPull only the first lever and your builders make apps that can't touch real data. Pull only the second and you have a smarter chatbot. Pull both and it compounds: people who can build, connected to systems worth building on.\n\nSame license. Same spend. What changed is what people can do with it. \n\n## Augment the human AND the AI\n\nEvery client conversation starts the same way now: \"We rolled out Claude. What's the ROI? Our spend is climbing.\"\n\nBoards are reading the stat that 95% of company AI projects show no impact on the bottom line. Sounds damning. That's the extinction rate. The question is whether you're set up to find your 5.\n\nThe usage numbers look bad everywhere too. 70% of the Fortune 500 pays for Microsoft's AI assistant. Maybe a third of those paid licenses actually get used in a given week.\n\nBut usage was always the wrong thing to measure. It only tells you whether people are chatting with the AI. The value shows up when people BUILD with it.\n\nThe answer I keep giving clients is two levers, pulled together:\n\n1. Augment the human: Give people a safe way to build, not just chat. Template repos with safety rails and automatic checks built in, trainings, a clear path from idea to finished app. Requests get triaged automatically. Roughly 90% get approved on the spot, and IT only has to look at the 10% that actually needs a human judgment call\n\n1. Augment the AI: That one guarded door connecting Claude to your internal systems. Logins, activity logs, usage limits, and permissions all handled in one place. Now the AI's answers come back with your own company data in them, and the tools your people build can reach something real\n\nPull only the first lever and your builders make apps that can't touch real data. Pull only the second and you have a smarter chatbot. Pull both and it compounds: people who can build, connected to systems worth building on.\n\nSame license. Same spend. What changed is what people can do with it.\n\n## The part engineers will hate\n\nHere's where I might lose some engineering friends.\n\nTell a senior engineer that people who have never coded are building real software, and watch them wince (hell I used to). And they have a point. A vibe coded app is not going to run your core systems. It is not enterprise software, and it shouldn't be.\n\nBut that isnt my claim.\n\nThe claim is smaller, and more radical: the salesperson builds the exact tool her team needs. The analyst turns the report she rebuilds by hand every week into something that just runs. Useful software, for a handful of people, that would never have been worth a real engineering project. That is buildable now. And useful beats impressive every time\n\nThat unglamorous layer is most of what we do at Tenex. Smart people thinking full time about one question: how does IT say yes without betting the company on it.\n\n## Intrapreneurs\n\nThink about what a business actually is: broken processes held together by people who care.\n\nEveryone shits on their own workflows. \"This part of the business is broken.\" Yea... Every business is broken somewhere.\n\nFor decades the deal was: file a ticket, get in the roadmap queue, wait. The people who understood the problem best had no way to fix it themselves. Their expertise was non-technical, so it stayed trapped.\n\nThat deal is ending.\n\nThe employees have the expertise and now they have a way to bring it to life. They can see a problem, think \"I can fix that myself,\" and actually do it. Intrapreneurs, with the tools to act like it.\n\nAnd if you're not the exec in this story, you're the leverage. The next load-bearing app at your company will probably be built by someone who has never written code. There's no rule that says it isn't you.\n\nThree years from now, the enterprise that pulled both levers has hundreds of small apps. Most dead. A handful load-bearing. A workforce that fixes its own problems.\n\nThe enterprise that locked it down is more secure on paper. And it missed the explosion.\n\nBut don't miss the fine print: the explosion happens either way. Your people are already building. The only thing you choose is whether it happens where you can see it.\n\nWe're at the start. Build the guardrails, open the door.\n\nI'm building this playbook (scaffolds, gateways, the SDLC for this new age) with enterprise clients at [tenex.co](http://tenex.co/). If you're staring down the \"we rolled out AI, now what\" question, that's what we do."},"adhxContext":{"savedByCount":1,"publicTags":[],"previewUrl":"https://adhx.com/AshwinKadaru/status/2072713807462281510"}}