Linotype Pilgrim

An early advent calendar in hard mode! Nice work

👍

Actually boost your career with this one weird trick Sam Altman hates:

"Learn the fundamentals."

—Abdullah Ijaz, Lahore

it's hilarious really because if one works 'bottom up', looking at the little problems computers can solve every day, the stakes for real, positive disruption are so incredibly low. tiny improvements add up.

the vanity and hubris of those at the top is absurd; they worsen every problem by breathing

the word (like 'ai', too) has become unusably tainted by them

"If [Israel] is treating Greta Thunberg this way, imagine how they are treating [Palestinian] women and children."

A lot of us never took that mantra seriously, because it's anti-engineering. You're correct that this whole era of 'vibe' (fake) coding is contiguous with it.

Startupism, disruption, all that bullshit has enabled the planet-sized fraud that Altman, Zuck, Huang etc are perpetrating on the rest of us

"well, ,you see Claude is a kind of mentor—"

Question asked of junior devs:

"How exactly do you plan to vibe code your way to senior level?"

#aifraud

well yes, but geographic might help sell 😇

Screenshot of LinkedIn posts by Abdullah IjazAbdullah Ijaz • 3rd+Verified • 3rd+ Software Engineer | Yapping about dev and career wins/fails 3w • 3 weeks ago • Visible to anyone on or off LinkedIn Claude Code has made me a dumber engineer. In the race of "optimizing" productivity, AI tools have slowly impacted my critical thinking and confidence as a developer. Last month, I was working on a personal project without access to Claude Code on my laptop. Suddenly, I felt lost. I'd pause after writing a function signature, expecting AI to fill in the implementation. Basic syntax that I knew by heart a year ago? I had to Google it. The worst part wasn't the slower pace. It was realizing how dependent I'd become. I used to have this gut instinct about code. That feeling when something's architecturally wrong, or when a particular approach just "fits" better. Years of debugging issues at 3 AM taught me to trust that intuition. But AI has been quietly eroding that. 1/3

Now I second-guess myself constantly. Instead of thinking through problems, I outsource decisions to Claude. It's easier, sure. But I'm losing the very skills that make experienced developers valuable. Here's what worries me most: Junior devs who are 90%+ dependent on AI these days - how exactly do you plan to vibe code your way to senior level? What happens when you get that panicked Slack message: "Website works fine locally, but production is down. Nothing in the logs." Will your response be "Sorry, Claude can't figure it out. I'll try different prompts tomorrow"? I'm making a change. From now on, I'm keeping AI completely separate from my editor. When I do use Claude, it's manual copy-paste only - after I actually understand the solution. Yes, it's slower. Yes, it requires more effort. But here's the thing - coding isn't just about shipping fast. It's about loving what you do and staying competent at it. 2/3

My advice to junior developers: Don't let AI do all the heavy lifting. Learn the fundamentals. Build that intuition through practice, not prompts. If you can't code without AI, you simply can't code. Your future senior self will thank you for putting in the real work now. 3/3

In conclusion, LinkedIn is a land of contrasts,

#aifraud

My new $261 billion facility will turn lead into gold, shame about that farmland and water and climate change but it will be worth it in the end, trust me

I am once again asking why these data centers are all fucking SECRETS

"A controversial plan to build a 2.2-million-square-foot data center on farmland may still happen. A rural community... is looking to settle a lawsuit recently filed by a development firm working for an unidentified tech giant."

bsky.app/profile/symb...

"This is why LLMs are pretty good coding agents"

is not the takeaway I had here.

You know who is? A fucking experienced human

Seriously. It's lodged in my mind from the very moment I first saw that tweet.

"Moving the slider" is something @edzitron.com writes about a lot. Because they have to. It's a constant game of the con men balancing their desperation to capture market share with the staggering losses they're incurring. But one thing is for sure: Users can never, ever win this one.

bsky.app/profile/symb...

Post on linkedin: Dave SlutzkinDave Slutzkin • 1stPremium • 1st Working on a new AI startup. Exited founder, builder, investor.Working on a new AI startup. Exited founder, builder, investor. 1h • 1 hour ago • Visible to anyone on or off LinkedIn AI products and agents make mistakes all the time, so patterns exist to architect around that. The screenshot is codex struggling to get something right for me. It writes bad code, sees the error, writes more bad code, sees the error again... eventually gets it right. But that's the key - it does get it right! (And yes, burns a lot of tokens doing it, we'll talk about this.) So here's the point about architecture: - Give the LLM a tight feedback loop -> it can self-heal. This is why LLMs are pretty good coding agents. If they can see an error, they just keep trying plausible solutions until they get past it. Not quite 10 million monkeys with typewriters, but maybe 10 highly-trained monkeys with text editors. But this is also why LLMs (still) aren't great coding agents. Not every change has such a tight feedback loop as this - lots require a user to click through and test, or something to be deployed to production infra, or a human to look at the UI and see if it's still broken. These are the cases where using an LLM gets really frustrating. The 10 highly-trained monkeys are ready to write the next iteration of code but they have to wait to get the feedback. 1/3

So there's a general lesson here for AI products. Your AI product is going to be more successful if you can have a tight feedback loop. - In coding agents it's the compiler/test output. - With ChatGPT it's the little thumbs up/thumbs down buttons - but no-one presses them - or the AB response test you sometimes get. - With Google search it's whether you click a link, go to more pages of results, rather than just getting what you need from the AI summary. - etc etc etc Build that feedback loop into the core product, it's really hard to tack on later. But tokens. The constant retries are going to hurt your token budget. So one of the interesting skills in AI product management is balancing correctness/retries with tokens/cost. It's a slider, like everything in product management. There's no right answer here - it depends on the stage of your product: 2/3

- If you're trying to nail PMF then eat the tokens. - If you're post-PMF and you're trying to make it profitable, you need to actively move the slider. This is also why you see AI products that work well for a while then stop working so well. They've moved the slider! 3/3

Every post about 'ai' coding 'assistants' is like 'yeah it's a huge time waster but this one time it worked,'

This is definitely gambling adjacent, and also maybe 'lizard corner' effect

#aifraud

Damn. This is potentially on par with the guy who discovered his therapist was using AI to counsel him. What a shit thing to do.

Another great quote from this talk:

bsky.app/profile/symb...

Kessler Kessler Text is the natural companion to Kessler Display, reflecting the same structure, but with the technical nuance required of smaller scale, long-form texts

Wow, Kessler

productiontype.com/font/kessler

Fans of Linotype Pilgrim (the real one, not me) will like this
#typography

Like I'm sorry, but the AI weirdos have DELIBERATELY created this confusion 100% so they can capitalize on the assumptions people who don't understand the difference will make.

This was the endgame of all the AI bros! no one knowing the difference!

Sorry they created this mess BUT THEY DID!

Really need people to understand what Ford is

OH "I'm always fascinated by mushrooms that look like futuristic moon bases in varying levels of disrepair"

— @penryu.bsky.social

And of course BlueSky has deleted Samaa’s account again. Her internet is too weak right now for her to create another one, so she asked me to share and let yall know. @mommunism.bsky.social @lexialex.bsky.social

gofund.me/db56c7ce3

I was already following Orb but a lot of people might wanna too

A someone who has written extensively about the eugenic roots that permeates those who claim to be building so-called artificial general intelligence, I'm here to tell you to please NOT compare criticism against corporations claiming to build a machine god, with eugenics. The audacity.

Let's see what we learn from this case of "Big corporation buys up important open source infrastructure because it's organized as a company".
#arduino

that's what the ceo said a few days ago yeah