6 min read
Let us talk about the elephant in the room.
Everyone is talking about Claude, about Mythos, about what it means for software engineering. My LinkedIn feed is wall-to-wall hot takes. Conference panels are being assembled right now with titles like “What Mythos Means for Your Engineering Organization” and speakers who have not shipped production code since the Obama administration.
This is not about Claude. This is not about Mythos. This is about the rate of change in capability that nobody in your leadership team is actually processing.
One year ago we were trying to get AI agents to build FizzBuzz in a single shot. That was the benchmark. Today I can build a complete platform, run it through tests, deploy it, and have it serving production traffic by end of day. I have done this. Not as a demo. As working software handling real requests from real users.
Twelve months. From “can an agent write FizzBuzz” to “an agent just built, tested, and deployed your backend.” That is not a linear improvement curve. That is a capability explosion, and it is happening while your organization is still debating whether developers should be “allowed” to use Copilot.
The models and the tools are going to keep getting better. That is not the variable you should be worrying about. The variable you should be worrying about is whether your organization can ship. As the tooling removes the technical friction from software delivery, what remains is the organizational friction, the stuff you have been ignoring for a decade. That friction is about to become the only thing that matters.
Your deployment pipeline takes four days because three teams have to sign off and nobody trusts the test suite because nobody invested in it. Your architecture review board has a six-week backlog. Your teams cannot ship a feature without coordinating across four Jira boards, two Slack channels, and a shared Google Doc that everyone edits and nobody owns. Do the napkin math with me. Fifty engineers, one major release per quarter, four days lost to approval theater per release. That is 16 days per year per release train of pure organizational waste. Multiply that by your fully loaded engineering cost and tell me you cannot find the budget to fix your test suite.
These are problems you should have solved ten years ago. You did not. The clock is running out.
I have a friend who was excited to take a VP role at a software organization. The mandate was exactly what he wanted: lead AI-driven development, modernize the org. Then he started his diligence.
QA was running manual regression suites that took two weeks. No test automation. A release process that involved a physical sign-off sheet (I am not making this up) that got walked to a VP’s desk. The developers were operating like it was 2015, not because they were bad engineers, but because the organization had never invested in the tooling or the culture to let them operate any other way.
He did the math. Before he could build agent-driven workflows or deploy AI-native toolchains or do any of the work he was hired to do, he was looking at eighteen to twenty-four months of remedial political pain. Convincing leadership to fund test automation. Fighting procurement. Unwinding a decade of process debt. And even if he won every one of those fights, even if he dragged the org from 2015 to 2024 in two years, by the time he got there 2024 would already be out of date. He would have spent his entire tenure running to stand still.
He turned down the role.
Your organization is now too far behind for good leaders to even consider joining. That is not a talent problem. That is a reputation problem that no recruiter can solve.
Your senior engineers making $200k-plus are doing the same math. Three weeks of PTO in the first year. Three weeks. In 2026. At a company that wants to compete for talent against orgs offering unlimited PTO, remote flexibility, and an actual budget for the tools their engineers need.
Speaking of budgets, your engineers are using AI tools on their personal accounts, paying out of pocket, because your procurement process takes six months and your security review takes three. Your organization never gave your security team a fast-path procurement lane for vetted AI tools. Your CISO is not the villain here. Your CISO was never given the mandate to solve this quickly. The engineers paying out of pocket are not just a morale problem, they are a data residency incident and an IP ownership dispute waiting to happen. Everyone loses. Unserious companies lose serious engineers.
Here is what happens next. Your board members are going to hire people to set up their OpenClaw accounts, the ones that generate a customized daily newsletter based on everything they care about. They are going to read those newsletters every morning, not skim, read. And those newsletters are going to tell them exactly where 2026 is. The cycle time reductions their competitors are shipping. The headcount efficiency. The time to market. Then they are going to walk into your next board meeting, and they are going to wonder why you are still in 2014. Rightfully.
You will not have a good answer. The real answer is that you spent the last five years managing around dysfunction instead of fixing it. You ran a 24-month transformation that is on month 36. You hired consultants to write maturity assessments instead of hiring engineers to write code. The board will not accept “we are transforming” when they can see the transformation is not producing results. They do not care about your SAFe implementation or your center of excellence. They care about one thing: is technology driving business outcomes, or is it a cost center that keeps asking for more money and more time.
So here is what you do. You find your best developers, the ones already using AI on their own time, the ones who have been waiting for permission that never came. You draft them. You give them a real feature to ship, not a proof of concept, not an innovation lab experiment, a feature your customers need. And you let them ship it.
Then you show up to their daily standup. Not to manage them. To map out the process. Every handoff, every approval gate, every place where the work stops moving and sits in a queue waiting for someone’s signature. Normally you would call this a value stream map and spend six months on it with consultants and sticky notes. You do not have six months. You do not have years. You have right now.
Remove the blockers as you find them. Not next quarter. That week. Fund the token budget. Fix the test suite. Kill the sign-off sheet. Every day that team is shipping, you are building the only evidence that matters: proof that your organization can move at the speed the tools now allow.
Because here is the thing you need to understand about the next twelve months. When that next model drops, and it will, the board is not going to ask you to evaluate it. They are not going to give you a quarter to run a pilot. They are going to ask why you cannot embrace it immediately. They are going to expect you to deploy it the way your competitors deployed the last one, fast, with confidence, into production. And if your org is still running manual regression suites and walking sign-off sheets to a VP’s desk, you are not going to have an answer. You are going to be standing on a stage you cannot get off of, explaining why the technology works everywhere except inside your building.
I do not think you can recover from that conversation. Not with your board. Not with your best engineers. Not with the leaders you are trying to recruit who already turned you down once.
Do you want to walk into that board meeting with a 24-month transformation slide, or with shipping software and a team that proved what is possible when you got out of their way?
