Michael Nygard

Author of Release It!, speaker

[Michael Nygard @mtnygard]

@cmuratori It's an LLM "tell".

[Michael Nygard @mtnygard]

@GergelyOrosz Another thing Github did was put the identity of the person before the identity of the project. Subtle but generates ego attachment which is a powerful motivator.

[Michael Nygard @mtnygard]

Every internal “replace Splunk” project I’ve seen woefully underinvested in the front end experience. The back end scaling challenge appeals to the engineers more.

Between AI to create a better experience and AI _being_ the experience I bet those projects will be a lot more successful now.

[Michael Nygard @mtnygard]

@valigo I agree. But vendors haven't seen returns on small conferences, so it takes a huge (media-produced) conference with a large expo hall & a lot of attendees before most vendors will participate.

[Michael Nygard @mtnygard]

@valigo But the sponsorships have dropped dramatically and the vendors can’t sell to developers (who only want to use open source). So the finances of conferences are super-risky for organizers.

[Michael Nygard @mtnygard]

@valigo The economics of conferences are brutal and force the organizers into a lot of financial risk. Especially the hotel space which requires early commitment before any tickets are sold. To compensate, conferences used to have big sponsor lists and expo halls.

[Michael Nygard @mtnygard]

@skylermzx I had that with GrapheneOS

[Michael Nygard @mtnygard]

@GergelyOrosz Sharma came from CoreAI, no?

[Michael Nygard @mtnygard]

Behind every increment of enshittification you’ll find a product manager hitting their OKR.

[Michael Nygard @mtnygard]

I kind of miss 140 characters. Don't need so many ambush newsletter posts.

[Michael Nygard @mtnygard]

@HazelAppleyard As you wish.

[Michael Nygard @mtnygard]

@Jonathan_Blow @ThePrimeagen I had the same thought.

This is a cross between “House of Leaves” and Nebraska Furniture Mart.

[Gene Kim @RealGeneKim]

What if organizations defined the developer role so narrowly that they actually make developers feel less capable, less confident they can solve problems, and more vulnerable from being automated by AI?

I was talking with Annie Vella, Distinguished Engineer at Westpac NZ, who has researched hundreds of developers across many different organizations. Her masters thesis might hint that the above might actually be happening. In the old days, developers did everything — they talked to customers, wrote code, deployed it, watched customers to make it better.

But for decades, many large enterprises have continued to shrink developer role down to just writing code. Architects make Visio diagrams, testers test, BA talk to customers and write requirements documents, PMs manage the project.

And the developer's job scope has been whittled to the single most automatable task.

Dr. Andrej Karpathy says AI is good at replacing tasks, not roles. But what happens when organizations have reduced a role to essentially one task (typing out code)?

Annie’s longitudinal research found that self-efficacy (not gender, not seniority, not experience) was the single strongest predictor of perceived productivity and positive developer experiences with AI tools. Self-efficacy is fancy talk for the feeling you have that you can successfully accomplish something: "I've done hard things before, and I can do hard things again."

Annie and I kept coming back to a troubling question: if a developer has spent a decade only being allowed to type code — never designing, never talking to customers, never deploying — where would they get the confidence to suddenly do more? Self-efficacy comes from mastery experiences.

What if the system they're working within that has narrowed their role also destroyed their ability to outgrow it? And the leaders who grew up in these hyper-specialized systems may themselves lack the breadth of experience to even envision what a broader developer role looks like. Or imagine what AI can do for their organization besides "automate coding."

The pinnacle of this madness: make the job codes in your organization hyper-rigid so that you can outsource the whole function, and switch between outsourcers every couple of years. The specialization is literally codified in contracts, making it even harder to undo.

In contrast, when developers know their job is to solve problems, they see AI as a superpower, not an existential threat. These are the developers who already know how to talk to customers, design systems, test their own work, and deploy to production. They don't need permission to do more — they've been doing more their whole career.

Learn how leaders are reshaping their organizations with AI at the upcoming Enterprise AI Summit, from OpenAI, Amazon, Disney, Block, Netflix, John Deere, Vanguard, and more. Full program below — see you there!

[Michael Nygard @mtnygard]

@nightwriter22 Villeneuve is such a visual director. I'm excited to see what he does with such a setting as Rama!

[Michael Nygard @mtnygard]

People often confuse description for endorsement and prediction for support.

[Michael Nygard @mtnygard]

@jessmartin Its pricing model was definitely built assuming much, much lower volume. I wonder if they can scale the infrastructure enough without massively hiking prices.

[Michael Nygard @mtnygard]

The audio versions of The Lord of the Rings, The Hobbit, and The Silmarillion performed by Andy Serkis are an absolute treasure. He really brings out the music in Tolkien's words.

[Michael Nygard @mtnygard]

@tlberglund @swimflythrive I'm not talking about discourtesy as much as the style of performance feedback. With humans I try for growth and development on strengths. That would be pointless with an agent. "Growth" in one session just gets lost with the context window.

[Michael Nygard @mtnygard]

I'd better make sure I don't start talking to humans the way I address my agents.