Eryk Orłowski · hretheumhretheum.com/questions/results-and-limits
Question P7 of 8
All questions7What did they deliver, and where are the limits?
At Wprost Online, after their advertising revenue strategy for 2003 was implemented, advertising revenue grew by 150%. The UX chapter built at ING kept working after leaving in 2022, with the same rituals, standards and career ladders. At Kokos.pl, reordered tasks around user goals, and at Morizon a loan application was led by an advisor in the form of a conversation; both solutions went into production and worked for clients. The limits are stated directly: the ING Allow Me concept did not reach production in that form, and Cognitive Mirror had no real users.source sources: wr-05, ing-12, mk-01, mk-02, mk-03, am-04, cm-08
Evidence
At Wprost Online I developed the advertising revenue strategy for 2003, after which advertising revenue grew by 150%.”
Both solutions went into production and worked for the clients.”
Gatefold produced a payment confirmation screen in SwiftUI, accepted as a code package, with all 30 VoiceOver checks passing.”
Rituals, standards, career ladders and people in their roles kept working after I left in 2022.”
At Kokos.pl I persuaded the client to reorder the tasks around the user's goals: getting a loan or investing.”
At Morizon the loan application was led by an adviser in the form of a conversation in natural language.”
The concept did not go into production in this form; ING implemented a permissions management module developed and imposed globally.”
The measure of the chapter working was that the same problem was not being solved by more than one team.”
A new company in the group is one configuration file, with no code changes.”
The studies made it possible to find and fix design problems early and to run a further iteration of the prototype.”
12 patterns accepted in Gatefold became the course on this site's job-posting subdomains, in production since 10 October 2026.”
Halvern is a signed and notarised pre-release for Macs with Apple Silicon, with a public repository under the MIT licence.”
Measured on the real model, the chat gave no facts from outside the package, fell back in 3.3% of answers and took 9.5 s for the slowest 5% of answers.”
Splitting the analysis into a separate call per dimension raised the small model's agreement from 20 to 25 of 26 points.”
In the paraphrase test the production model gave the same result in 39 of 40 cases.”
Cognitive Mirror has had no real users, and its instrument is not empirically validated.”
A measurement on the live stack showed that 5 repetitions of the same prompt gave 5 byte-wise different results.”
I designed the "agent in the product" route and suspended it on the same day, before implementation.”
As of 30 September 2026 no hypothesis has a result yet.”
Honestly, there were too many to pick one, and most of them had the same shape: I'd start the next five things before finishing the first. […]Honestly, there were too many to pick one, and most of them had the same shape: I'd start the next five things before finishing the first. Only recently did I learn that I have ADHD, and that explained a lot of it. The difficulty with finishing tasks wasn't a lack of discipline, and to-do lists were never going to fix it. Getting diagnosed and getting professional help changed how I work. So did Exocortex, the knowledge system I built for myself, which today is a real part of how I function. If you recognise yourself in this, get a diagnosis. It helps you understand what's going on, and with the right support you can actually manage it.”
Komitywa (K2 Internet), the company where I was managing director and a partner between 2007 and 2009. […]Komitywa (K2 Internet), the company where I was managing director and a partner between 2007 and 2009. It didn't survive, and looking back the reasons were mostly mine. We went at it on instinct, without a properly thought-through strategy, and that works about as well as you'd expect. I also learned that I'm not necessarily the best salesperson, which is a useful thing to know about yourself when you're the one who's supposed to be selling. Both lessons shaped how I work now: I plan before I build, and I'm honest about which roles I should leave to someone better at them.”
Nothing went wrong, actually. […]Nothing went wrong, actually. Hireverse was my 2025 experiment with a reverse interview: instead of only answering a company's questions, I let it explore my background on its own terms. It worked, and people enjoyed it. But a good joke is only funny for the first hundred tellings, so once it had done its job I froze it on purpose rather than keep it alive for the sake of it. Knowing when something is finished is a skill too.”
Limit
I take too much on myself instead of delegating. […]I take too much on myself instead of delegating. When something matters and I know how to do it, the fastest route always looks like doing it myself, and for a single task it often is. Over time it's the wrong trade: I become the bottleneck, and people in the team don't get the chance to grow into the work I'm holding on to. I've learned to catch the moment I reach for something, and to ask whether it really has to be me or whether I'm just faster at it today. Most of the time the honest answer is the second one.”
E-mail: [email protected] (kt-01) · LinkedIn: linkedin.com/in/eofek (kt-02)