ljm

There aren't that many good ways to mentor or share experience in a team without pairing though. Sometimes it's just better to sit down with someone and directly collab, and let the person with less experience do the driving so they get the muscle memory. This is an easy transfer of skill and knowledge and it works well because it's hands on.

AI doesn't make that very easy if you're just two people staring at an agent, but I think a lot of people who aren't already senior are going to hit a wall they can't climb when their experience limits what they are capable of doing with AI.

As with many things, usually the bad experiences with collaborative methods come from a specific implementation and those experiences become the bias.

Because, like it or not, it does Write Once, Use Anywhere better than Java ever did.

It is pretty much the lowest common denominator for code.

You're calling it ass covering but I kinda see it as buck-passing too. Seems like D'addario didn't want to hold the buck so they kept passing it down until it made its way back to the worker who did it.

Seems to me like that worker's ass wasn't covered at all and now everyone's all up in his shit on his socials, and the layers of management get to wash their hands, even though it's the management and executive that are pushing AI at all costs.

So they're not human then. The quote holds up.

It won't be made of people for long given how many Ellison is culling to prop up his gamble on AI data centres.

It amuses me that you have consistently refused to draw the connection to this 'ocean' of asphalt and the purpose of the building it was designed for and are basically being obtuse.

But you know what is a better neighbour than a data centre or an office or a strip mall? A nice public park and local shops that support the community without you having to drive your car a distance that in any other country would be easily walkable.

Why would anybody want a monolithic megastructure sat next to their house that offers nothing good to the community itself? There is no upside for the locals. No improvement to quality of life.

I'm failing to understand what is so impressive.

The credits at the start are off centre, biased to the right hand side.

The prose that follows is fully centred, so it hangs to the left of the actor names, and it has a little bit of rich formatting. It's not even good prose because it is overloaded with emdashes, almost as if it's direction. People will say this is like AI but even AI has more restraint than this.

So the fucking amazing title card for blade runner is a studio intern scrolling the first couple of pages of the screenplay through an overhead projector. Because it follows the exact format of a screenplay.

But really it is the audio that is doing all the work. That is ominous and spooky and futuristic as hell.

Nothing ever improves in the tech scene... it just decays until someone challenges it.

Funnily enough CF has redesigned its navigation several times in the past year, just like GitLab used to do. Shuffling the UI like a deck of cards is pretty desperate.

The announcement reads like it's generated by AI. I opened it looking for a punch list of selling points but there was nothing. I figured it was just Cloudflare trying to do Cowork.

The readme I generated for an agent I put on top of pi reads almost the same as this.

AI does not do emotion. There is no punch, or pop. If I walk away from an announcement thinking it's AI slop then the announcement failed.

It's nothing like the posts Cloudflare used to write.

Bring back FastCGI, enable the setting in Apache to list a directory, and point the agent to your cgi-bin directory full of all your tools. Calling the bin without any input provides the help text or the interface to the tool.

You can probably throw together a passible auth system on top of it that works with the htaccess format. Which is another cgi script but chowned so it can't be called from the outside. If not that, public key crypto - give your agents certificates that are part of the chain of trust. Mint those out of band and give them an insanely short TTL. Fail2ban or something like that to prevent abuse (agent getting handsy with the server gets blocked, credentials revoked).

The side benefit is that everybody starts writing perl and awk again. Who needs a JS runtime when you have CGI?

It's still too early stage for me. Anthropic needs to mature for a little longer before they're enlightened enough to understand my kafka-based event sourcing model for cross-agent orchestration. Kafka, protobuf, functional domain modelling using F# as a DSL...the whole kit and kaboodle.

Only when my AI and its subagents are modelled through a graph of producing agents, consuming agents, and agentic mailboxes, will Claude Code truly shine.

Probably see it as the easy way out because actually bringing about change in a culture that is already entrenched is hard work and they don't have the mettle for it. Or don't trust the org to be on board with it.

Easier to just do a reset instead of thinking about how one might bring effective change over a reasonable period of time.

At risk of sounding like a grump, once I started at a company where almost everyone is partnered up and has kids, and almost everyone basically has nothing to talk about except their partners and kids, I had the perfect incentive to stay away from almost all social events. Or stick around for an hour at most because of how boring it is.

It became pretty clear that drinking culture in a lot of places develops around people having a lot in common, and certainly becomes worse when one of those things in common is 'enjoying drinking'.

I don't mind this. I had all of my 20s and most of my 30s for drinking after work. Now it's not really something I care about.

I'm amused by Claude developing its own corporate jargon which, while clearly pulled from the corpus of human language, is quite distinct from the bullshit traded over the boardroom table.

I've had trouble unpacking some of the copy it generates because its default tone of voice is fairly unintuitive. It's almost like it's designed to induce psychosis because it makes you feel like you've discovered something novel.

Of course I have. If I see ghostty constantly consuming the most energy and helping reduce my battery life to barely 2 or 3 hours, the first thing I'm going to do is switch terminal to see if I can improve that situation. The built in terminal and WezTerm have been fine and I've had much more reasonable battery life since.

This does not even speak to it pegging my CPU and keeping the fans running on full blast even when the laptop is supposed to be idle overnight.

> bun was vibe coded by a single dude without any open source community involvement.

I think even this is largely beside the point. The acquisition of any project by Big Capital (whether tech or VC) is usually a killing blow.

I never knew that running an interactive program in my terminal would absolutely rinse my CPU and battery but that's what Claude, OpenCode and Ghostty have colluded to achieve. Even when the laptop is asleep overnight it's practically melting.

I'm sure there was some logical reason for shoehorning web technology into this stack given that we have a good 40 years or so of experience with interactive terminal programs that use curses/ncurses, alongside emacs, vim, almost the entirety of MS-DOS, and so on.

How does that work out in terms of usage? I imagine that, unless you're on the highest tier subscription or have a no expense spared approach to agentic coding, then you're going to hit a wall pretty fast.

It's very strange to vibe code a project and feel both a lack of emotional investment but also the cognitive overhead of steering an AI.

The process of iteration, or the feedback loop, typically allows space to try things out and experiment and then either drop an idea or refine it more purposefully. And those breaks in between the pure delivery to focus on that gives you room to breathe and also see the progress of your work.

Blatting it out in 6 hours with Fable 5 skips all of that and you have something like an MVP, but you haven't really put anything of yourself into it. So why bother committing unless you can take that further or apply something novel or reflect your own personality in it? Or you genuinely believe you're on to something and the AI approach actually gave you a path to building it?

I've been 'vibe coding' something, I've spent about a month on it on and off along with a fair bit of debugging on some sticky issues. I still think I'd suffer going into that codebase and doing stuff in it by hand, no matter how well I thought I organised it all.

Anthropic's suite of products have practically zero cohesion across each other. I've said it some time before.

There is no clear product vision or overarching roadmap in terms of users and UX, it is just vaguely connected things being vibe-coded in parallel.

I figured pi itself would be the best harness because it's barebones and you make it what you want. omp is to pi what doom is to emacs is what lazyvim is to neovim.

Purely anecdotally the one persistent issue I have with LLMs writing code is that they are absolutely paranoid and add a load of indirection and defensive crap and even if you prompt to avoid that it will often require manual steering to remove the cruft.

> Pi is meant for people who know what they are doing

How many people genuinely know what they're doing when the value prop of Pi is basically to vibe code it to your taste? The entire point of vibe coding being that you don't actually have to know what you're doing?

Nothing in (AGENTS|CLAUDE).md, no Cursor style rules. Just some skills for Jira/Confluence, CI, and Playwright.

Haven't bothered doing any extra customisation in Claude Code or Codex as I don't really trust those things to apply consistently. And if you can run your CI steps locally before you push you don't need to tell an agent to remember it.

That said I've been playing with pi today and given how stripped down it is I've used it as a sandbox for customisation. Still, I haven't gone quite as far as laying down project level instructions. More that Claude Code in particular is quite heavy, and so it its prompt, and adding stuff like rich integration with GitHub and SourceHut (CI status, active branch or PR) seems comparatively trivial in pi.

Only thing left to do is switch out ghostty for the time being as it's misbehaving. Laptop hot to the touch even after idling overnight.

Needed to pay my bills. It tided me over for a short while.

Waiting for the next gig. This was before I learned that sometimes it's better to take the hit and walk away.

I've struggled with this, even encountering people who basically say "if AI can do it why do I need to spend any more time?"

It was disappointing hearing someone tank their own prospect of career growth like that.

They'll frame it as not wanting to waste anybody's time by submitting you as a candidate if you're already mid-flight on other positions, but they can still figure that out without having specifics on the actual company, role, salary, etc.

They're banking on you offering it because saying no would rule out any mysterious prospects they have to offer, but really they're looking for new leads and if it comes from a candidate then it's warmer. Just have to be polite when you give them vague info and say you're not able to share more than that.

I had a really weird power-trip interview once, and it was the CEO who was joined by a contractor who was ostensibly the technical lead for the project. There was nobody in engineering who worked full-time on the job, everybody was contracted. I was also going in as a contractor so I was aware that the interview would be a bit lighter on process as a result.

The questioning very quickly veered away from technical stuff and into stuff like, "where do you stand spiritually?" and other questions probing into whatever bizarre cosmic insights I could pull out of my ass at the time. He was the really intense kind of boss who wants to make sure you know of it with the hard back/shoulder slaps and micromanagement, and I could see his office from the boardroom which basically had an array of monitors all wired up to CCTV so he could watch (and hear) people from the comfort of his desk.

If any of that wasn't a red flag, getting hired literally 5 minutes after leaving the office was probably the biggest. I lasted about 6 months and even trying to leave was an ordeal.

It kind of is though. Someone else will say "why are you sourcing results from an Israeli company?" and another will say "why are you sourcing results from a Chinese company?" and another one yet will say "why are you sourcing results from the US?".

Why are the ethics of working with Yandex or Baidu any better or worse than the ethics of working with Google or Microsoft? Except that they're not western.

The logical answer is that a person like this wants a very strong firewall, so ethically impure bits don't cross into their LAN.