Ages ago, someone described the "squashed middle" of management to me. That level where your boss throws stuff down and expects it to be dealt with, and the level below you expects care and emotional availability. It stuck with me. On a day when I was feeling a bit too much like a punching bag, I went for a walk - it was raining, how apt - and thought about it. How hard it is to give what you don't get. How it can start to seem like being the punching bag is the job, and how miserable that can be.

I was, I guess, a middle manager, in an organisation that did not like that word. I reported to the CTO. I had nine managers underneath me. I think that is the definition - a manager above you, managers below. How limited, a definition of a place on the org chart, and not what it is you actually do.

The recent round of AI-justified layoffs has been clarifying about what executives and VCs actually thought middle managers were doing. The Block post - Jack Dorsey and Roelof Botha, two thousand years of military history landing on "there is no need for a permanent middle management layer" - lives rent free in my head. The role is information transfer, no more no less. Coming from remote environments, this blows my mind - though I guess in organisations that rely on meetings to transfer information, that inefficiency made the routing role necessary. Where I remain confused: why is it AI that changed this, and not moving away from hand-delivered paper and pigeons as mechanisms for transferring information?

Which brings me to my core question: how many managers were that great? I have many times interviewed managers and written in my notes that they seem to see the job as operating their position on the org chart. Perhaps I was being too harsh; that was the definition, what they were expected to do. It's hard to be great working from a definition that was flawed and lacking at its premise.

The Block post does preserve a role for developing people - we're calling this "player-coach", apparently. Barely an improvement. I keep seeing these words, I have yet to see any kind of definition of what they actually mean.

In the ZIRP era, the metric was headcount, and the value of a manager as an abstraction of people. Now the metric is token spend - equally flawed. Both vanity metrics around input, not around impact. No wonder we don't have a good definition of what a great EM does. And no wonder people feel thrown for a loop - they were never anchored on the right metric to start with, and have little to hold onto when that changes.

I'm more than half way through LSE's MBA essentials, feeling like I'm getting an inside look at how capitalism reasons. If speed is no longer a moat, output is, then the differentiator is human leverage. Companies that figure that out will have a structural productivity advantage, and that is why this is playing out so aggressively. This is why the decisions that seem wildly risky are happening anyway - because the win, if it comes, will be so large.

The thing I find most fascinating, and think about near daily, is CircleCI's 2026 State of Software Delivery report finding that top teams nearly doubled throughput year-on-year while median teams barely moved. The differentiator was delivery infrastructure and discipline, not AI tooling. Existing team practices are the groundwork that turns adoption into productivity gains.

Team practices. Strong process. Delivery. Also known as: what great managers do.

Despite my critiques of the flawed reasoning, I do start to find myself sympathetic to replacing workers with AI. This is how I'm scaling myself right now - instead of nine direct reports, I have nine terminals. Claude has expanded beyond editor and developer to intern recruiter, program manager(s), content manager, content strategist, a place where I dump in the thoughts as they pass by and have it organise them. It's pretty adequate at most of these things, because I took the time to define them, build the underlying data it needed, segment the tasks so it can work within them. The work of being a manager, essentially. The work I've done for a long time. The work some people seem more willing to do for a robot than they were for a human being.

We need to get to a better definition of manager. Jean and I spent weeks debating this before launching The EM Survival Guide. We landed on an overarching definition of an EM is a force multiplier for the team, and then broke down how, tactically, that can be achieved. It's been great to see it land with our first cohort, to get the feedback that people now understand what their job should be.

For myself, a bit over three months into a new life, what's strange is how much has changed and how much hasn't. The mindset I come in with each morning is unchanged: my orientation remains how I can make everything around me better today. The actual work is very different. I had barely accepted I was the IT department at Twill, and now I am also the infra department - I ventured into AWS settings for the first time this week. The most important thing for DRI right now is marketing, not something I know much about, and the next thing I need to go deeper on. Process and efficiency are still the core of what I do. They are just abstracted by skill.md files now, instead of by people.

This time is hard. And somehow I'm still optimistic. I think it's because I would always rather have the truth. When the gloves are off, you find out what people actually thought the job was. Uncomfortable, but useful. The first step in strategy should always be understanding the context you're in.

What I'm doing

• Wrapping up the first cohort of the EM Survival Guide. I really love working with leaders who are so committed to supporting their teams, and happy to see the format pushing them to take some time for themselves and reflect. The next cohort starts in June.

• Working on our latest course Navigating the AI Shift, which starts May 11. Jean and I built it for those engineers and EMs who are not Extremely Online bragging about workflows, who want space and support to work through what this all means for them.

• We just kicked off the second cohort of DRI Your Career, with three full scholarships this round, which I'm proud of. Really love seeing people working through the shoulds capitalism has installed in them and thinking about what they want to choose instead. This runs again in July; we have super early bird pricing for a limited time.

• Continuing to learn about the inner workings of capitalism in LSE's MBA Essentials. Latest modules have been accounting, which I always thought was just... accounting for things. Turns out it's also part of how decisions get made.

• Continuing to be a fractional CTO at Twill. A lot of effort this month has gone into the AWS migration. Mildly terrified but excited about all of it!

• ELC Conference in Prague was wonderful. My favourite talk was Natalia Nazaruk on debugging confidence - wrote about it on Mastodon. Also really enjoyed a couple of extra days in the city with my partner afterwards - restorative.

• Preparing for the Engineering Shift Conference in June - online, organised by Findy in Japan. I'll be doing "What's My Job Again? Developing Self-Management."

• I have new postcards based on What Raccoon Are You. Ping me an address if you want one.

What I've been reading

Powerfully Likeable by Dr Kate Mason. I picked this up because Jean recommended it, and wished I'd had it years earlier. The chapter on imposter syndrome did not pull any punches:

"The only thing wrong with imposter syndrome is all the wrong people have it."

Mason reframes it as "imposing syndrome" - the worry about taking up space, which limits what you let yourself reach for. There's a lot of practical structure in the book too - her 15-minute prep model is gold.

Steve Yegge's "Vibe Maintainer" - it's a hard time to be an OSS maintainer. Yegge proposes an alternative to the no-AI-PRs stance: triage with agents, fix contributors' code yourself, cherry-pick the good parts, make rejection the last resort. The token cost to do this at scale seems very high - my continual complaint about many of the most bleeding-edge AI workflows. But the implication is interesting: if maintainers are going to spend tokens reconstructing your contribution anyway, the highest-value thing a contributor can do is write a really good issue. With output abundant, the spec is more valuable than the artifact. Critical thinking is the scarce resource.

The feedback leaders stop getting - we all know feedback is important, but how often do you see someone admit what they're actually working on? I loved the honesty of this piece. "How would you introduce me to someone who doesn't know me?" - such a good question to take with you.

Frank Warren and the art of small confessions - some of my favourite art is this shape. PostSecret. The Museum of Broken Relationships in Zagreb. Individually the pieces don't meet the bar of something that would be in a museum. Collectively they do. There's something about pulling together a myriad of perspectives from single snapshots.

Know someone who'd enjoy this? Forward it on - or they can subscribe at whatsmyjob.club 🦝