DEV Community: Jonathan Blessing

Every Day is Y2K

Jonathan Blessing — Wed, 08 Apr 2026 16:22:27 +0000

Twenty-six years later, I think we should remember Y2K —a crisis named after its deadline.

My first job in New York was as the BBC's Y2K coordinator for the Americas. I spent months preparing systems for a single night. Fix the code, test and re-test the systems, and hold my breath at midnight. The sun rose the next morning and Y2K was happily forgotten.

Mythos-class AI does not offer a date. It offers a condition. Thousands of zero-day vulnerabilities surfaced in weeks, with no reason to believe the rate slows. The next model will find more. The one after that, more still. There will be no midnight.

Mythos-class AI does not offer a date. It offers a condition.

Consider: Mythos found a 27-year-old zero-day in OpenBSD —the operating system built, from the ground up, by people whose entire reputation rests on not having exploitable code. It is like discovering that if you hold water just so, it explodes. If that system had a door no one found for 27 years, what is hiding in the systems that were not built with security as a founding obsession?

This week, tech companies privately briefed the White House on what Mythos means for national security. The conversation is no longer theoretical.

Imagine a Y2K-level event every day. We are moving into a permanent state of crisis.

Most people have not yet absorbed this because they are still thinking in terms of events —a breach here, a hack there, each one reported, investigated, and forgotten. That is the old model. The new model is not a series of events. It is a climate.

The new model is not a series of events. It is a climate.

This is the new operating environment. Not a crisis to be resolved, but a tempo to be endured. We survived Y2K by treating it as an engineering problem with a deadline. The institutions that survive this will be the ones that recognize there is no deadline. The present does not end. It only compounds.

The question is no longer whether the vulnerabilities exist. Mythos answered that. The question is whether the rate of discovery will outpace the rate of repair. For any system where theft is irreversible —crypto being the most obvious, but not the only one— the math is not encouraging.

So much for the forecasted employment crisis. What comes instead is a permanent mobilization. The work of hardening systems, triaging vulnerabilities, and patching what each successive model discovers does not end. AI will not replace the workforce. It will redirect it —into an unending cycle of repair. The machines will not take your job. They will make sure you never finish it.

Originally published at launchdayadvisors.com

Mythos is Where Crypto Ends

Jonathan Blessing — Wed, 08 Apr 2026 12:29:10 +0000

I have been waiting for the thing that comes after Opus. Here is my prediction: Crypto ends here.

Imagine keeping your hard currency on the front lawn. No dog. No police. Thanks to Claude Mythos —the thing that finds every unlocked door on the street— this is now the state of digital assets. Unlike hard currency, which must be moved through time and space, across borders of protection —friction rooted in physics— crypto can be taken instantly and irreversibly.

Here is the thesis: you do not even need to break into the house. Near-chain exploits —vulnerabilities in the browsers, operating systems, and infrastructure that people use to access their wallets— will be enough to enable irreversible theft at scale. A model that can surface thousands of zero-days in weeks changes that equation overnight.

The weakest surfaces are not the blockchains. They are the wallet apps on unpatched phones, the DNS providers, the browser extensions, the SMS codes from telcos still running ancient signaling protocols. These are not the walls. They are the lawn. And there is no fence. New vectors of attack will emerge in these layers, and they will not be hardened in time for the widespread availability of Mythos or its many successors.

Anthropic did something unusual: it built the most powerful AI model it has ever created, and decided against a widespread release. But as Benjamin Franklin observed, three can keep a secret if two of them are dead. The capability might not be widespread, but it is out.

In just weeks of internal testing, Mythos identified thousands of zero-day vulnerabilities across every major operating system and web browser —including one that had gone undetected for 27 years. Every one of those vulnerabilities is a door. Some of them lead to wallets.

The disruption to crypto has not arrived. That is not the same as saying it will not. Take a minute to consider what else risks being moved to the front lawn.

Originally published at launchdayadvisors.com

Faster Than You Think: AI Will Hit Like an Asteroid

Jonathan Blessing — Thu, 26 Mar 2026 18:36:16 +0000

Every major technology has followed the same arc: big promise, slow delivery. Wake me when it's over. ERP promised transformation and delivered years of implementation. Cloud promised speed and delivered migration programs measured in years. The grander the claim, the longer the integration cycle.

The pattern is predictable. Value doesn't appear until systems are rewired. APIs are mapped. Data models are reconciled. Machines have to learn to speak to machines — through hard-won, handwritten code — before enterprises benefit.

Skepticism about AI's speed of adoption is reasonable.

And yet this cycle will move very differently — not because AI is smarter, but because of where it's injected into the enterprise. Previous technologies demanded system-to-system integration first. ERP had to replace accounting workflows. CRM had to connect to sales infrastructure. Data warehouses needed pipelines before a single dashboard could roll out.

AI doesn't start at the core. It starts at the human layer, on our side of the monitor.

Like an eager new hire, it works with what's already deployed. It reads the documents teams already read. It drafts inside tools they already use. It summarizes the reports executives already review and reasons over exported data without redesigning the database that produced it. No middleware. No migration. No eighteen-month integration roadmap before the first win.

That difference is what makes this one move like an asteroid — fast, and visible only once it's already close.

AI carries the biggest promise yet, but its first layer of leverage doesn't require ripping apart the architecture. You can generate material gains before the plumbing is complete. Durable advantage will still require data discipline, governance, and workflow redesign — but you don't need to finish that work to start.

That is why this will be the fastest adoption cycle ever seen.

Not because enterprises are suddenly agile.
But because AI begins where every other technology ended — at the human layer, on our side of the monitor.

If you're making decisions about AI implementation partners, our guide to selecting an AI development partner can help you navigate the vendor landscape.

Part 2: Be the Asteroid

Part 3: Instant Dashboard

Originally published at launchdayadvisors.com

How Claude hit our Agency Vendor Like an Asteroid

Jonathan Blessing — Thu, 26 Mar 2026 11:37:25 +0000

A case in point: this website — and how Claude hit our agency vendor like an asteroid.

Six weeks ago, Liz and I retained a UK-based design & development agency. While demanding, we were ideal clients (so we all think). Before kickoff we shared the copy, the logo, and a fully formed design system. Their mandate was straightforward: design and development. Ideally in a month.

A month and a week later, we were still circulating homepage comps in Figma.

I've experimented with most of the LLMs — technical tasks, non-technical tasks, large and small. As recently as six months ago, my bias was simple: I'd rather pay a professional than wrestle with the tools myself when it comes to coding.

A nerd friend and heretofore LLM skeptic had been urging me to spend time with Anthropic's newest release since it launched. So on the afternoon of February 16th, I decided to dip my toes into Claude Opus 4.6. The goal was modest: push the homepage forward with a clickable prototype.

I fed Claude our design system, homepage copy, and a project brief. Boom.

As I was about to ship the results to our agency, log out, and sit down for dinner, Claude asked whether I wanted to complete the rest of the site. Within hours, I had roughly 80% of what you're looking at now.

It went into production that night.

We fired the agency in the morning.

AI didn't need to integrate into the agency's workflow. It didn't need a migration plan. It operated at the human layer — on my side of the monitor — and collapsed a six-week project into an evening.

If you're evaluating AI partners or wondering how to structure an AI implementation, our guide to selecting an AI development partner covers what to look for.

Be the rock, not the dinosaurs.

Part 1: AI Will Hit Like an Asteroid

Part 3: Instant Dashboard

Originally published at launchdayadvisors.com

India’s $315B AI Survival Thesis

Jonathan Blessing — Wed, 25 Mar 2026 02:09:43 +0000

The Economist published a piece this week examining why AI has not yet disrupted India's IT outsourcing industry, a sector they treat as representative of the global market's exposure to AI displacement. The conclusion: legacy code is messy, clients overestimate AI's readiness, headcount keeps growing, and Nasscom expects its members to post combined revenue north of $315 billion this year. Crisis averted.

The analysis is not wrong. But it mistakes a lagging scorecard for a forward indicator.

The Brownfield Defense

The strongest argument from India's IT executives is the brownfield one. The Economist quotes Atul Soneja, Tech Mahindra's COO, distinguishing between greenfield environments (new systems with clean architecture, where AI excels) and brownfield ones, where legacy code, missing documentation, and interdependent services make AI deployment far harder. His argument, essentially: AI works great on a blank canvas, but enterprise reality is never a blank canvas.

Fair enough. Anyone who has tried knows the difference. And knows he's correct.

But the defense assumes AI capabilities are static. They are not. The reason brownfield environments resist AI today is that the tools cannot hold enough of the system in view at once. They lose track of how services connect, where the undocumented dependencies live, how a change in one module cascades through twelve others. That limitation is dissolving. The context windows that govern how much code and documentation an AI model can reason over simultaneously have expanded dramatically in the past year alone. In practical terms, an AI that could previously read and reason over a few dozen pages of code can now process the equivalent of several thousand pages in a single pass. Agentic tooling, AI that can navigate codebases, run tests, and trace dependencies autonomously, is maturing in parallel.

The moat around brownfield complexity is real today. It is also eroding, month by month, release by release. And "yet" is carrying enormous weight in a $315 billion survival thesis.

Revenue Is a Lagging Indicator

The Economist cites slightly-better-than-expected quarterly results and rising headcount as evidence of resilience. But aggregate revenue growth reflects contracts signed twelve to eighteen months ago. It does not tell you what is being signed today. As we wrote in Vendors Under AI Pressure, when revenue compresses, margins tighten, and the pressure shows up in delivery long before it shows up in earnings.

The Consulting Pivot and Its Limits

The piece quotes Nandan Nilekani, one of Infosys's founders, projecting that AI-related services could be worth $300 to $400 billion by 2030. The instinct is right. Companies need help deploying AI effectively. Understanding organizational context, business logic, integration constraints: that is consulting work. Humans remain better at it. This is the strategic layer where experience and incentive awareness still matter more than automation.

The limitation is scale. India's IT industry employs roughly 5.4 million people. The vast majority are not performing strategic consulting. They are writing routine code, maintaining test suites, handling support tickets, processing data. The strategic consulting play can preserve the industry's margins. It cannot preserve its headcount.

This is the distinction the optimistic narrative elides. The services that survive AI pressure are not the services that employ the most people. For buyers navigating this shift, the question becomes how to select an AI partner that understands these structural dynamics, not just the technology.

Context Windows Will Prove This Thesis Wrong

The Economist closes by observing that AI's effect on the sector remains "unclear and uneven." That framing lets you be right regardless of what happens next.

Here is a more specific claim: context window expansion alone will unravel the brownfield defense. When an AI agent can ingest an entire legacy codebase, every module, every configuration file, every undocumented dependency, and reason over it coherently, the argument that these systems are too messy for automation stops holding. That capability is not speculative. It is the current trajectory, and it is accelerating. We have been writing about this speed since we launched.

The industry will not collapse. It will compress. When one skilled developer with the right tooling begins doing the work of several, labor arbitrage narrows. Not overnight. Gradually, and then structurally.

The disruption has not arrived. That is not the same as saying it will not.

Photo of brown field by the author, traveling through Spain with his dad.

I Needed a Dashboard. Claude Built It in an Hour.

Jonathan Blessing — Sat, 21 Mar 2026 17:28:09 +0000

Our site runs on Cloudflare. For infrastructure at scale —DNS, caching, edge routing— they're exceptional.

Their analytics dashboards, however, are a 747 flight deck. I needed a Honda Civic dashboard.

We don't use cookies on this site — by design. But my cofounder Liz and I were curious what pages people were hitting and how traffic was flowing. Cloudflare provides analytics, of course. But accessing them means living inside their overwrought dashboards — clicking through panels designed for enterprise infrastructure teams, not two people who want to know which blog post got read yesterday.

So I asked Claude to generate a lightweight analytics view directly from our HTTP logs.

A simple /analytics endpoint. Basic stats. Clean display. Secured behind Cloudflare's zero-trust authentication.

No wireframes. No middleware. No integration roadmap. No sprint planning.

It's just done.

What I Actually Did

I gave Claude three things:

The Cloudflare Workers environment the site already runs on
Access to the HTTP request logs available at the edge
A plain-English description of what I wanted to see: top pages, traffic over time, referrers, status codes

Claude wrote a Worker that reads request data, aggregates it in memory, and renders a minimal dashboard behind Cloudflare Access. No external dependencies. No database. No third-party analytics script injected into the frontend.

The whole thing — from prompt to production — took about an hour. Most of that was me deciding what I actually wanted to look at.

Why This Matters

This isn't a story about dashboards. It's a story about where AI operates.

Every previous generation of technology required system-to-system integration before you saw value. ERP had to replace your accounting workflows. CRM had to connect to your sales infrastructure. Data warehouses needed pipelines before a single chart could render.

AI doesn't start at the core. It starts at the human layer — on your side of the monitor.

It works with what's already deployed. It reads what you read. It builds with the tools you already have running. No middleware. No migration. No eighteen-month integration roadmap before the first win.

I didn't need to file a ticket, wait for a vendor's sprint cycle, or sit through a scoping call. I described what I wanted in plain language and shipped it the same afternoon.

The Velocity Is the Point

Six weeks before this, we'd retained a design agency to build our site. A month into the engagement, we were still circulating homepage comps. I gave Claude (Opus 4.6) the same brief one evening and the site went into production that night.

The dashboard was the same pattern, compressed further. An hour instead of an evening.

This is what adoption velocity looks like when the technology doesn't need to integrate with anything. It just needs a person with a clear idea of what they want.

That's the thesis: AI will move faster than any technology cycle before it — not because enterprises are suddenly agile, but because AI begins where every other technology ended. At the human layer. On our side of the glass.

Be the asteroid.

I'm Jonathan Blessing, cofounder of Launch Day Advisors. We're a buyer-side advisory firm that helps organizations select AI, software, and UX partners. If you're evaluating vendors for a technical engagement, we run the search, diligence, and negotiation so the outcome reflects your interests — not theirs.