Toby Rogers

surprised to learn that the file format .docx is actually named after an English novelist.

popularity is not a measure of quality. just because something sells a load of units, it doesn't mean its better than everything else.

it’s much harder to build something genuinely simple that solves one problem well than something with a ton of features.

someone unsubscribed from my newsletter after i posted about muzak. i didn't think there were any real muzak fans out there.

while you port and emulate games, I ported Jira to my PSP

don’t just read more books—do something with what you learn from them.

AI denialist travels back in time to kill Claude Shannon, only to return to the future where all software development is done with "Alan Code"

its Bacchus this is 100% how he wants to be respected

creating reams of text about a subject often just means you haven’t really developed your ideas about it—even more so if you just get an llm to do it.

jobs-to-be-done is still one of the best frameworks for figuring out what you should build. designing for the circumstance your customer finds themselves in is much more useful than designing for a persona.

ai writing is the muzak of prose

five is better than four

It’s really fucked up that Channel 4 would put someone’s private key in their logo like that tbh

I’ve switched from spotify to apple music because 1) less nazis, 2) better sound quality and 3) book club radio dj mixes. all worth it

controversial, but halloween iii: season of the witch is my favourite movie in the halloween franchise (after rob zombie’s halloween 2)

i’m that certain kind of person

The chickens are coming home to roost after decades of the software industry treating security as an afterthought. Anthropic’s warnings of a software security apocalypse from AI models like Mythos now seem less like hype and more like sober warnings.

How bad is it? This bad:

today is brought to you by…

in the ai slop era when anyone can build anything, restraint is a product management superpower.

a pm with enough experience can make subconscious judgements about what’s good or bad in their domain. that’s what product taste is. it’s what john boyd called “fingerspitzengefühl” and it’s not something you can shortcut.

the beatles didn't learn their craft in a rehearsal room—they learnt it in hamburg nightclubs, playing for hours to drunk sailors who'd walk out the moment they got bored. where does a junior pm get an audience like that now?

at its heart, a product manager’s job is to transform ideas into winning products. ai doesn’t change that.

Is taste really a moat? For something to become a moat it has to be defensible. What’s stopping LLMs getting “good enough” at taste that the human element becomes irrelevant?

Super-interesting interview with Sam Altman on the newly-rebranded Every podcast.

one person unsubscribed from my newsletter after this week's post landed. they clearly don't like the velvet underground.

this week's punk pm newsletter: is your product ahead of its time, or is it just not very good?

the best thing about ai models getting better is how much less scaffolding you need around them. i'm just running single chats for most of the things i'm working on now instead of messing around with projects and folders and .md instruction files. opus 5.5 just does what i want.

This episode of Lenny’s podcast with Peter Sellis (ex-Snapchat and Discord) is a fantastic listen. Loved his comparison between great product teams and terrorist organisations. There’s definitely some truth to that analogy.

every day i imagine there’s an extremely high-stakes meeting inside google between the forces of “we can’t keep showing these dogshit ai results” and “doing anything else would cost 20 trillion dollars”

Building the tool stopped being the expensive part. Knowing what you want it to do never got cheaper, and that is the step most people skip even on the settings screen. The ones who will build their own are already the ones hunting for workarounds.

Frontier Psychiatrist sounds like an Anthropic job posting

This week’s newsletter is about “lost genius” Lee Mavers from The La’s, and why perfectionism is really just an excuse for not putting your work in front of people.

Hyper-personalisation with AI is very cool, but I don’t think everyone is going to start building their own personal software tools anytime soon. I mean, most people never touch the default settings on the apps they already have. How much is AI really going to change that?

These new models are cool and everything, but can they log into Microsoft Teams first time?

Remember when people used to make fun of AI for not being able to count the number of R's in strawberry? Now whose laughing ITS FUCKING ASTRA THATS WHO

There are some rumors, some allegations, that I got the chosis. The AI psychosis. Let me just put those to bed: Do I use agentic coding for everything now? Yes Do I think LLMs can be useful for a ton of great things? Yes I have the chosis yes that’s what I’m getting at

gotta love avowed marxists who *checks notes* opposes a specific technology full stop instead of opposing how it's being used by the modern-day factory owners to extract further surplus value from the modern-day factory workers like i'm sure karl would have agreed with that, right?

I do not negotiate with kids who don’t want to do their homework.

Talking to PMs at companies embracing smaller AI-native pods, it’s creating demand for senior PMs. AI lets smaller teams build more things in parallel, which creates more decisions about what to build and how it all fits together. AI reduces the cost of building and increases the value of judgment

“Yeah, that sounds great. But does anyone actually want it? Maybe we should find that out first.”

Hashing things out with an LLM works not because the LLM magically has all the answers, but because this process forces you to structure and reflect on your thoughts in a way that helps you find the answer yourself. Software developers worth their salt know this as "rubber ducking"

I’m not going to believe in AGI until there’s a model that can figure out PlayStation’s parental controls.

Non developers seem to have absolutely no awareness that everything important is coded with AI now and theres just no escaping using it.

If LLMs never improved another iota for the rest of our lives, it would still be the most transformative invention of the 21st century. but it keeps improving

One thing I have learned talking to lots of people about AI is that they can be both worried about the implications of AI and very excited about using AI themselves. I think people on this site tend to view many people's attitudes to AI as much less complicated than they are.

This is how Skynet got started. Some guy called Kyle told me.

The biggest problem with the product builder archetype is you end up with product people who don’t know anything about engineering cutting code, and developers who’ve never spoken to a real customer shipping apps.

Curation and sensemaking are the new content creation. When our feeds and inboxes are flooded with AI-generated “content” we need people who can help us find the signal in the noise—and help us figure out what it all means.

Taste as a moat has become a bit of a mantra lately, particularly in conversations about the impact of AI on product management and design. This post from Cong Wang is interesting, though, because it comes at if from a software engineering perspective.