miketerhar.com
Some sort of devops or system or problem refinery. Smooth out everything but make sure you have all your dysfunctions together in one spot that gets scrutiny.
392 posts
146 followers
225 following
Getting Started
Active Commenter
comment in response to
post
*Daisy-like: remember the great gatsby? Where daisy came to town, killed a lady, gatsby took the blame and got shot in his pool? (Spoilers)
comment in response to
post
The biggest pro I’ve seen is that it takes a long time between starting the outsourcing effort and many executives can get promoted and move onto new jobs elsewhere before the wheels fall off.
I’ve seen this played out on a few healthcare and investment systems and the Daisy-like execs keep going.
comment in response to
post
Gave jippidy write the release notes based on the code changes:
* adds surveillance for monetization
* adds bugs to several functions
* allocates way more memory for no reason
comment in response to
post
I always like the way Maine people navigate. “You can’t get there from here.” Is a thing in Maine. You have to go somewhere else first, then you can get where you’re going.
comment in response to
post
What are you using to kick them off? Are the Claude things running in actions or on your device?
comment in response to
post
Part 2: oath enforcers.
Anyone who took an oath to the constitution gets raked over the coals if they undermined it or ignored court orders (checks and balances are in there!)
Expel any congressperson who betrayed their oaths. The constitution is important.
comment in response to
post
From the blurb I sense he’s differentiating Slips from other kinds of errors. Very good.
comment in response to
post
Whew. I was just very concerned that talking of Private Link was about to get me canceled but this doesn’t sound anything like that thread.
comment in response to
post
But it’s just Fannie Mae and the Fed doing a little dance. In the 1980s, they’d eat up loans for like 200k so houses all went up at that price point. That moved up to 300k in the 1990s because interest went down. 400k by 2000, 575k by 2010… so no developer would ever target a lower price point.
comment in response to
post
Is Atlassian a team that is to be emulated? Their software is some of the worst I’ve been forced to use. If going to the office once a quarter results in that, let’s do something different.
comment in response to
post
This is huge. I can point to lots of stuff about early GitLab but the culture alignment was very strong because the values were tied directly to cash spot bonuses. For the low low price of $20,000, you could give 20 people $1,000 spot bonus every month. cash changes behavior like nothing else!
comment in response to
post
Socializing isn’t saying something and inviting 20,000,000 people to scrutinize it. Socializing is sitting around a table or bar or sport thing or camp fire with fewer than 10 people so if you say something awful, you get made fun of and correct your takes in a safe environment.
comment in response to
post
I’ve only ever run Ruby on Rails at scale and the “convention” approach that makes everything magically line up is a huge liability. It’s nice to be able to assume things will work but there are simply too many ways to do things so if you’re not DHH and also the architect, anything can break.
comment in response to
post
Rails always felt to me like the first step toward vibe coding. You could run a cli to make more controllers and pages and actions. Your db stuff would just kinda exist after a bit of suggesting what activerecord should do.
Great for prototypes. Terrible for production.
comment in response to
post
I would NOT keep a poo-covered chandelier. Very unsanitary. Difficult to scrub.
comment in response to
post
They’re coming for our toilet paper archives now. Fml
comment in response to
post
I’m not a fan of the auto-lint correcting thing. I want it to get the thing working, then lint, then instrument, then unit test.
comment in response to
post
Is this just “bullshit producers benefit most from bullshit machine?”
As a bullshit producer, I do feel that it benefits me with its bullshit firehose.
comment in response to
post
With Hugo, accessibility is in the theme or it’s not. Finding a theme that works is the only thing that may be difficult if you don’t want to spend time with go template mustache nonsense.
I think the LLMs are pretty good at it though.
comment in response to
post
Did we learn nothing from Pacific Rim?!
comment in response to
post
I think that’s why everyone fell in love with Tailwind. You can just guess style names and they kinda work and that’s that. None of this cross-component coordination bullshit.
comment in response to
post
Sadly it’s proprietary so they won’t share and it’s contracted “by the hour” so their only deliverable is timesheets which sucks. But some people are trying.
The evidence that it is beneficial is that there’s no control group. No team wouldn’t use AI at all.
comment in response to
post
One consultancy I know of is testing it on a task they do a lot of. A specific kind of transformative engagement where multiple teams are working with multiple customers in parallel.
They have different teams working with different IDE and LLM combinations and are checking the results.
comment in response to
post
It’s really easy to plagiarize coursework so, of course, the plagiarism machine is good at it.
We can’t let the framing shift to a calculator. It’s not computing answers from abstraction.
If there is plagiarism-proof coursework, let’s use that. Otherwise, we need to ban it.
comment in response to
post
Is Microsoft careful about retaining enough staff that the 6,000 fired people can’t raise funds and disrupt them using their experience and skill? This seems like a big risk.
comment in response to
post
They also do stuff like block the .env file so if you’re not being a maniac with your api keys, even when an LLM tries to manipulate the .env file, it can’t do so.
comment in response to
post
There’s definitely a feeling that the folks you’re talking about maybe don’t have a good sense of their own consciousness and how it relates to human intelligence.
Hell, I’m not sure if I could prove my executive functioning is much more than a transformer model that has meat controls for my limbs
comment in response to
post
Unpack it a bit for us?
comment in response to
post
Why “go to work anyway”?
That seems a bit bossy for a robot that is supposed to be uninterested. I’d have preferred the answer that it doesn’t care if you want to or not but also it doesn’t care if you go or not.
comment in response to
post
It does remind me of a former boss who got into security in the 1990s because his name was an SQL injection name that started with “O” and an apostrophe.
Control signals being inline with data is an unending source of insanity. And we baked it directly into every LLM.
comment in response to
post
I’m gonna be the best damn junior developer ever! Half million for my wondrous trash code.
comment in response to
post
I have been using GitLab pages for years and will not stop! It has redirects and authentication if desired.
GitHub pages is probably fine but I am a gitlabber so their pipelines and things make more sense to me.
It can be a built app or just html also.
mjt.sh
mushtache.com
winnie.fun
comment in response to
post
I’m not a full time dev with a backlog so this may be different than most… but I find the barrier for getting started is much lower. I don’t have to have a really good idea for how it will work in the end to try a few things.
Since my coding is jammed between other tasks, this is hugely important.
comment in response to
post
I haven’t gone super deep on either but it looks the same as astronomy ~> astrology.
You take something observable and give it some supernatural meaning.
Planets moving or LLM output.
comment in response to
post
The one thing I find most interesting is that OpenAI and Anthropic seem to use their own indexes rather than google’s. Since Google is so awful, if you want to use these other indexes, you need to go in via Claude or ChatGPT and then check its references
comment in response to
post
Agenda is good. I loved when I was at GitLab it was standard practice to decline meetings where there wasn’t a clear agenda (including an outcome or decision). And leaving a meeting as soon as it was clear that you weren’t needed.
I don’t know if it’s still that way, but it was nice to have agency
comment in response to
post
The main concern being everyone who becomes a good editor spent time becoming a good writer.
Everyone who becomes a good senior developer wrote a lot of terrible code and had to deal with it for years.
Can people get good without the drudgery?
Electrician apprentices think not.
comment in response to
post
One thing I’m testing out with less experienced folks is if they can develop the same gut check as I have when the LLM provides info.
I don’t know every NPM module, but I know how to see if they’re legitimate before I install it. How much “open source ecosystem”-like stuff do people need?
comment in response to
post
A bakery near me makes Kouign-amanns and I’d never seen them before. I became “somewhat addicted” and have not been able go to a week without having one since I found them.
comment in response to
post
I’ve only ever seen success with compensating people a bit more and seeing them rise to the occasion. It’s insane that the manager class considers it their job to suppress wages as much as possible.
It’s sad that it requires a benevolent oligarchy corporate structure to do.
Unionize?
comment in response to
post
It’s about building your own in-house Heroku.
🙃