“The difficulty lies not so much in developing new ideas as in escaping from old ones.”
— John Maynard Keynes, The General Theory of Employment, Interest and Money

‍

‍

Last time, I wrote about the siren song of ineffective metrics, like lines of code converted, and how value should really be measured by the outcome: working software. Even on the way to that glorious outcome, lifting thine eyes from mere code conversion yields perhaps an even greater value: control in undertaking a highly uncertain task.

‍

Remember the movie, Speed, starring Keanu Reeves and Sandra Bullock? A disgruntled ex-bomb squad member decides to rig a city bus full of passengers with a bomb that will explode if the vehicle speed drops below 50 miles per hour. At one point, the good guys save the hostages by driving another bus alongside and putting a ramp between the two, allowing each passenger to step into safety (while tampering with the bad guy’s video feed, of course, so he can’t see what’s happening).

‍

Because I’m in that pitiable state in which I think about work all the time, this struck me as an apt analogy to explain how Mechanical Orchard moves workloads into production with minimal disruption. We provision a ready-to-go cloud-based environment next to a running mainframe system and, as each workload achieves equivalence, it moves over while both systems are running side-by-side.

‍

(The obvious difference is that people moving over are not newly built replicas of existing passengers... but hey, all analogies break down eventually.)

‍

This is what makes it safer, especially for systems that absolutely cannot stop working. It gives you a viable path to the goal of decommissioning the mainframe entirely—or in the case of Speed, letting it collide with a cargo plane and explode spectacularly—without impacting operations.

‍

What’s more, this method gives you control in a highly uncertain situation, something in precious short supply these days. Not the semblance of control, as outlined in Gantt charts and burn downs, or in those essentially useless Red/Yellow/Green “status dashboards” that remind me of a behavior chart in preschool. But actual control of small, measurable progress. Specifically:

‍

1. Because you’re not changing anything as it goes between old and new, it means you essentially click “Undo” if anything unexpected happens.
2. As each workload moves over, you gain confidence in how the rest of the system will progress: the fog of uncertainty dissipates the more you do, and you can see far more clearly what resources are required to get to the finish line.
3. You can show stakeholders what’s happening. Not what might happen, or what’s hypothesized to happen, but what’s actually happening. This is gold for your personal hoard of political capital.

‍

Put another way, when you’re able to reduce risk and uncertainty, you’re able to increase the predictability and trajectory of a difficult task full of known and unknown unknowns. In Speed, staying above 50 mph wasn’t the goal—it was a way to buy time until a real solution could be executed. Converting code without a broader strategy is motion without resolution: the real value comes when you build the ramp.

‍

‍

News and Views

Friends, this is what political capital looks like: “In practice, the middle managers who understand the nitty-gritty and control day-to-day implementation of projects hold the real authority.” The Economist incisively explains why the barriers to AI adoption aren’t technical at all: they’re predictably human (subscription required).

‍

I used to commute between Portland, OR and San Jose, CA, almost always on Alaska Airlines. Seems like they’re encountering some IT-related turbulence recently. This follows on an issue that took down a system during a modernization effort in April.

‍

The incentive structure is not looking promising. Google’s Gemini Deep Think took a gold medal at the International Math Olympiad. How can one possibly ask teenagers to study their trigonometry and integral calculus when all they need is an LLM? Asking for a friend.

‍

From the Orchard

Six is magic: we now have Very Large Customers across the banking, consumer packaged goods, retail (there’s even a case study), public sector, manufacturing, and automotive industries.

‍

We’re interested and somewhat interesting. Matt Wynne joined Tuple’s Jack Hannah about testing, pairing, and how “social programming” makes modernization safer. David Laing spoke with Ken from London Tech Talk about integrating LLMs into his coding workflow—comparing AI tools to PhD students. I chatted with James Barker on the HiveMind podcast about trust, clarity, and what really matters in B2B tech marketing—especially when AI and metrics start to distract. Roundup here.

‍

If you’re in London or time zones nearby, come see us at the Gartner Application Innovation and Business Solutions Summit at the Intercontinental O2 on September 8-9. Edward Hieatt, our CCO, will be speaking about “proof, not hope” and forwarding a better metric for measuring modernization progress: workloads in production.

‍

Curious to learn more? Say hello@mechanical-orchard.com.

‍

—

*Issue first published on July 29th, 2025. View all newsletters here

‍