ljm

Google doesn't need to give a shit, because so much of the internet is infested with with google ad trackers and adwords, and everybody uses Chrome, that they will continue to make billions even without AI. Facebook did the same with their pixel so they could soak up data.

Gemini will be dead in 2 years and there'll be something else, but the ad and search company will remain given that they basically own the world wide web.

Except now, so much of the WWW is filled with AI slop that it breaks the system.

Why are you thinking more about the people that piss you off than the ones that you consider hard working and caring?

You have a massive chip on your shoulder, dare I say that's why you've had many caring managers and now you're seeing them all as 'social climbers'.

Did one manager call you out on something and you torched the entire thing?

A smalltalk or Erlang for AI agents is an interesting thought. Smalltalk for the design in terms of message passing and object-oriented holding of state (agents are stateful and are reached via their public interfaces), Erlang for the elegant execution of it with actors and mailboxes (agents have inboxes and outboxes and can work concurrently at scale). Might as well go the whole hog and put a supervisor AI agent in as a switchboard.

UK has the state pension which comes from NI (National Insurance) contributions which in a way acts like a Defined Benefit pension in that you work for X number of years and get state pension of Y in return(currently adjusted for inflation, wage growth, or 2.5% annually, called the 'triple lock'). Not based on income so you don't get a massive state pension by earning 6+ figures.

Then more recently (as in, as of around 2012 and up to 2018) we got auto-enrollment into private pensions, which are more like Defined Contribution (DC). Employer has to pay a percentage into the pot and so do you. Usually 5% employee and 3% employer by default but many will offer better (or contribution matching) as a perk. I think this is probably the same or similar to the 401k in US terms. The employer chooses the pension provider but you need to proactively switch to a high risk scheme to see any growth from it.

At a certain point the tax rebate from the government doesn't cover your whole income so you have to file a tax return to get the rest of the rebate. You can instead choose to 'salary sacrifice' which means you are lowering your income on paper but the sacrifice is put directly into pension (or otherwise can be used to get a car on lease or a bicycle via another scheme). Salary sacrifice is used by a lot of higher earners to bring their gross income down in order to avoid being cut off on certain benefits like child-care.

After all that you have SIPP (Self-Invested Personal Pension) which gives you more control over what you can invest it. Not just stocks, ETFs, and all that, but can also be commercial property (so the pension itself owns that asset). This gets the same tax treatment for pensions.

Finally there is the ISA and LISA. The first is a savings account where any interest or capital gain is free of tax, the second gets a 25% boost by the government to help buy a house or flat, but you can only use that money for a mortgage deposit.

Most people won't see all that much from their auto-enrollment given they could just opt-out to get the extra cash in their paycheque (especially when a low earner), or might not know to switch to the high risk fund, so the state pension and other benefits for OAPs will be there still. Those with more disposable income or a long term view (e.g. doing FIRE) are likely to max out the various vehicles available to them but at that point you're gonna be earning too much to care.

No need to optimise healthcare costs or any of that unless you want to go private.

If you can't hold something down for more than a year then it looks rough to anybody.

18-24 months generally seems to have been the sweet spot in terms of improving your income. Especially because the promise of equity after 4 years often means fuck all and you almost never see a promotion or a decent pay increase either for inflation or performance.

If an employer is going to ding you for taking advantage of the market then they better be offering enough above market to keep you around for longer.

I'm not really sure how that's exclusive to software engineers either. But OP was drunk at the time of writing so I guess they were feeling a bit lovey-dovey.

I think the real underrated skill to learn as an engineer is how to test.

Documentation can take many forms: ADRs, system diagrams, specs, JIRA tickets, commit messages and descriptions, PR descriptions, code comments, idiomatic/intuitive code, etc. etc. and much of that requires maintenance or review to ensure it's still up to date.

Outdated tests quickly become broken tests, and they serve a purpose as documentation as well, but aside from throwing around buzzwords like TDD and BDD and all that, it's rarely a skill that is explicitly developed. Or maybe it's handed over to an SDET or something.

Build a decent set of tests that can survive the implementation being refactored, rather than coupling them to the runtime code or filling them with mocks and stubs, and you can get a good way to documenting complicated routines simply because you are explaining how it works through numerous well-described assertions.

That means you never need to bother with the 'how' or 'what' when commenting code, and you have multiple levels of 'why' as you go up from the code to commits to the issue tracker and beyond.

What company can't pay 36 euro and legislate it after?

Are hetzner in the wrong for denying service to clients in a deficit?

Not so much atrophy as apathy.

I've worked with people who will look at code they don't understand, say "llm says this", and express zero intention of learning something. Might even push back. Be proud of their ignorance.

It's like, why even review that PR in the first place if you don't even know what you're working with?

Mermaid has been great for a similar reason. For example, you can render a mermaid diagram inside a PR description on GitHub.

Comes in handy when describing a state machine or the flow of data.

I've also observed the third direction, which is the message storm.

That's when someone breaks down their point into multiple separate messages, one sentence at a time, when a single coherent paragraph or two would have worked. Why send one message when you can send 7 in rapid succession?

It's arguably the most annoying method of communication because it spams your notifications and you have no idea when someone has finished dumping.

There was a screenshot of Valve's front page back when Half Life 2 was released in the early 00s[0]. It was well laid out, straight to the point, and had design flourishes that would have been painful to put together at a time where CSS was new and not supported very well.

Obviously a product of its time and laid out similar to how it'd be printed in a magazine (the characters slightly overflowing the borders and such like). Accessibility wasn't a thing back then.

If a different company did that in 2018 you'd be seeing the G-man in corporate memphis, downloading about 500mb of assets, with 178 separate ad trackers in a consent popup, and then you'd be scrolling like mad to get through all sorts of animations that hijack the scrollbar, in order to get to any useful info.

[0] https://www.reddit.com/r/HalfLife/comments/10sx4ve/what_stea...

I do similar, but I dislike writing CSS because it's practically impossible to keep up with the standards. And because I dislike writing CSS I don't feel like writing HTML that much either.

Web Components were a bit too slow to take off so the mental model of JSX has stuck with me, even if the ecosystem with hooks and various approaches towards reactive state are in many ways inferior to a problem Smalltalk already solved back in the day.

They're also getting into cloud compute given you can use the desktop app to work in a temporary sandbox that they provision for you.

I was about to call it reselling but so many startups with their fingers in the tech startup pie offer containerised cloud compute akin to a loss leader. Harking back to the old days of buying clock time on a mainframe except you're getting it for free for a while.

AI can already generate music and it stands out a mile both in terms of the lyrical content and how middle of the road it all is. People will create some slop on Suno and lip-sync to it on TikTok because they absolutely don't sound like a thousand country singers blended into one.

So it's competent, for sure, but that is damning it with faint praise.

We can skip Web3... Web 4.0 is twilight gradients, glassmorphism, text size xs in tailwind, and cards and pills for every UI component. Along with self-explanatory help text acting as filler under every header.

It's a legitimate boon for throwing together little internal apps, mind.

I can slap something together with Claude over a few evenings to fill a gap on tooling, or I can wrestle with Jira and CI and all that to tie things together with their own integrations.

No thanks, I'll just take the API keys and build on top, to my exact specifications, and the interface will be passable even if it needs a lot of polish. Tailwind has worked wonders for that.

The remaining issue is that even an AI-generated UI needs considerable UX input in order to work well, especially when you have to fit it around domain specific knowledge, use-cases, and prior art. Is it for power users or not? All that.

At risk of shifting the goalposts on what I originally said, unique here isn't meant to mean quirky or weird but, simply, something that hasn't been done before, or hasn't been done as effectively.

This is the challenge for B2B startups that are switching to LLM-based development and are trying to offer more than the reselling of cloud compute at a markup with specialised functionality, because AI turns SaaS into a sexy version of MS Access.

I reckon something like this has only been possible to develop because of how homogenous the internet has become in terms of design ever since the glass effect and drop-shadows took over in Web 2.0 and Twitter Bootstrap entered the scene.

You'll get a competent UI with little effort but nothing truly unique or mind-blowing.

Impressive technology, but that old skool artisanal weirdness of yore only becomes more valuable and nostalgic.

Even a distributed or silo'd org chart has some affinity across the hierarchy in order to keep things in overall alignment. You wouldn't expect to use a product suite that is, holistically, not fully compatible with its own ecosystem, even down to not having a single concept of a project. Or requiring a CLI tool in an ephemeral environment that you cannot easily configure.

That's clearly a trade-off that Anthropic have accepted but it makes for a disappointing UX. Which is a shame because Claude Desktop could easily become a hands-off IDE if it nailed things down better.

I've noticed a lack of product cohesion in general and it does make me wonder if it's a result of dogfooding AI.

For example, chat, cowork and code have no overlap - projects created in one of the modes are not available in another and can't be shared.

As another example, using Claude with one of their hosted environments has a nice integration with GitHub on the desktop, but some of it also requires 'gh' to be installed and authenticated, and you don't have that available without configuring a workaround and sharing a PAT. It doesn't use the GH connector for everything. Switch to remote-control (ideal on Windows/WSL) or local and that deep integration is gone and you're back to prompting the model to commit and push and the UI isn't integrated the same.

Cowork will absolutely blow through your quota for one task but chat and code will give you much more breathing room.

Projects in Code are based on repos whereas in Chat and Cowork they are stateful entities. You can't attach a repo to a cowork project or attach external knowledge to a code project (and maybe you want that because creating a design doc or doing research isn't a programming task or whatever)

Use Claude Code on the CLI and you can't provide inline comments on a plan. There is a technical limitation there I suppose.

The desktop app is very nice and evolving but it's not a single coherent offering even within the same mode of operation. And I think that's something that is easy to do if you're getting AI to build shit in a silo.

I could tolerate Windows more if it stopped being a rental OS that thinks MS own your laptop more than you do. Like it was in Windows 7.

All I ask is that things I uninstall stay uninstalled -- I got rid of OneDrive and Teams for a reason, stop adding that shit back! -- and that it doesn't shove Edge and Bing down my throat and decide that MS knows better than I do about what I want.

I'm a huge fan of SCP, or at least early day SCP. The problem I've found with a number of the longer form sagas is that they tend to rely on world-ending apocalyptic events (which really does get boring, it's not high stakes any more) and are heavily anime-coded.

The authors do put in a great deal of effort, which is laudable, but it does make me wonder if the writing style and story telling is a deliberate choice, or if the authors simply watch anime more than anything else and thus the universe-saving power fantasy is the only thing they know how to write.

Personally I enjoy the more 'grounded' and mysterious stuff.

I heavily advocate for rawdogging AI agents.

All the fancy frameworks are vibe coded, so why could they do better than something you do by yourself?

At most get playwright MCP in so the agent can see the rendered output

Never have I seen a company try so damn hard to make something a thing than Microsoft and Copilot.

And it is absolute dogshit. And offensive to actual copilots.

Similarly Wether spoons, the chain of pubs in the UK.

More interestingly, they tend to set up in historically significant or listed buildings and as a result, preserve them. Not unusual to find a Spoons set up in an old 19th century bank or something.

Kotlin is practically a no-brainer when you have JVM at your finger tips, versus something like Swift which is comparatively young.

I tried to use Vapor with Swift recently and struggled to get something working because the documentation looked comprehensive, but had a lot of gaps. I ended up throwing it out because I didn't have the time to dig through the source to understand how to do something, when I could use a mature framework in any other language instead.

The promise is there but I'm just not ready to invest. My youthful days of unbounded curiosity are coming to an end and these days I just want to get something done without much faff.

You could get a pint of nothing but head as well, if you wanted.

In Spain they try to create the head by dumping the beer in the glass, Guinness style, and letting it excessively foam up. So it's half flat and half full.

I worry that CI just got overcomplicated by default when providers started rocking up with templated YAML and various abstractions over it to add dynamic behaviour, dependencies, and so on.

Perhaps mixing the CI with the CD made that worse because usually deployment and delivery has complexities of its own. Back in the day you'd probably use Jenkins for the delivery piece, and the E2E nightlies, and use something more lightweight for running your tests and linters.

For that part I feel like all you need, really, is to be able to run a suite of well structured shell scripts. Maybe if you're in git you follow its hooks convention to execute scripts in a directory named after the repo event or something. Forget about creating reusable 'actions' which depend on running untrusted code.

Provide some baked in utilities to help with reporting status, caching, saving junit files and what have you.

The only thing that remains is setting up a base image with all your tooling in it. Docker does that, and is probably the only bit where you'd have to accept relying on untrusted third parties, unless you can scan them and store your own cached version of it.

I make it sound simpler than it is but for some reason we accepted distributed YAML-based balls of mud for the system that is critical to deploying our code, that has unsupervised access to almost everything. And people are now hooking AI agents into it.

The real impressive examples get turned into SaaS prototypes and not placeholders for your imagination.

If they had vision they wouldn't be thrown out in a blog post.