I participated in a hackathon meant to introduce vibecoding to people! Wrote up my thoughts about it here.

This is an actual line that was added to the official system prompt for Codex for GPT-5.5 by OpenAI. Usually the system prompt is as minimal as possible, so I assume it would otherwise mention goblins a lot. AIs are weird.

Me previously: Waaaah I don't understand terminal everything is so hard and this is a personal lack on my part Me now: Is the default font size 11px and line height 1? That's literally unreadable for me, no wonder I was having trouble.

The claim isn't that social media is dead (yet). It's the assumptions that made it coherent as an object — human production, social circulation, public observability — can no longer be presupposed. We need new concepts and methods capable of studying what comes next.

(b) Semi-private spheres — group chats, invite-only communities, subscription spaces. Reciprocity over reach, trust over virality. Structurally invisible to most existing research methods. (c) AI-mediated communication — We today talk more to ChatGPT than we interact with friends on social media.

So "social media" is fragmenting into three post-social formations: (a) Algorithmic broadcasting platforms — few-to-many flows from creators, brands, and bots, delivered as personalized feeds. Broadcasting, but individualized. Endpoint: the fully synthetic feed.

Third: a retreat from public posting. A ~50% drop in posting on Facebook and X between 2020 and 2024 (see my previous thread). Ordinary public participation is hollowing out, while sustained conversation moves to WhatsApp, Discord, Signal, Substack.

Second: generative AI is breaking the link between platforms and human content production. Over 40% of long-form Facebook posts likely AI-generated by late 2024. More than half of long-form LinkedIn posts in 2025. "Slop" is the word of the year.

We are no longer commenting or even liking. Just watching. Likes, shares, comments — "active" signals — are being displaced by watch time, dwell time, scroll hesitation. Preference is inferred from micro-reactions. The 'prod-user' has become the user-spectator.

Three dynamics are now dissolving it. First: the algorithmic shift. Platforms moved from social graphs to interest-based recommendation. You see what the model predicts will hold your attention, not what your network shared. TikTok set the template; everyone copied it. The social network is dead.

For two decades, "social media" meant a specific setup: users produced content, content circulated through social ties, platforms extracted value from mediating this. Networked publics, digital trace data, platformization — a generation of research rested on that configuration.

Exhibit A: in a recent antitrust filing, Meta argued it isn't a social media monopoly because it isn't really a social media company anymore. Only 7% of Instagram time and 17% of Facebook time is spent on content from friends or followed accounts. The rest is algorithmic video from strangers.

Social media is no longer social. Most of it is passive viewing of videos and pictures from people we've never met. But we're still studying social media like it's 2010. We've entered the post-social media era — and research needs to catch up. osf.io/preprints/so... preprint w/ Richard Rogers

Interesting choice! The serif and muted colors of the logo are definitely pointing to a more formal, mature look. Still seeing a lot of the old, energetic red-orange in links and buttons though. Is that going to be updated later or will it be kept?

Blegh, government websites continue to be awful. I'm checking in on my local county's jury duty website, and it is slow to load and confusing in a way that turns what should be joyful civic duty into a source of stress.

I will admit, I am coming around to AI use on some things. Claude has been very helpful as a way to get direct instructions on solving technical issues, such as figuring out why I'm not getting notifications on my laptop or setting the clock on a specific microwave model.

🔋Grid-scale batteries are replacing gas plants in CA’s evening energy peak! On 3/29, CAISO batteries delivered 12.3 GW (43% of demand) during peak hours, sustaining over 20% of the grid for hours during the evening ramp CA’s battery capacity has surged from 1.3 GW in 2020 to ~17 GW todsy

If you can substitute "hungry ghost trapped in a jar" for "AI" in a sentence it's probably a valid use case for LLMs. Take "I have a bunch of hungry ghosts in jars, they mainly write SQL queries for me". Sure. Reasonable use case. "My girlfriend is a hungry ghost I trapped in a jar"? No. Deranged.

I'm churning these out here's my recipe 1) buy domain they like for 1 yr 2) make atproto account with their email 3) add TXT record for custom handle 4) add CNAME record for blento 5) link handle and blento custom domains 6) make a basic blento page 7) send them the atproto password

Saw the announcement of footnote implementations from @leaflet.pub and I am so excited, you have no idea. I'm going to port over all my fandom meta on my other account. Feedback in next post. lab.leaflet.pub

Dealing with the government of Canada's websites as a product designer is a uniquely infuriating experience. I didn't even know anyone still had text input fields with left aligned labels! This is awful!

You don’t even need that imo. A per se rule that the entity offering the LLM into commerce is vicariously and strictly liable for the LLM’s speech. It’s a narrow and coherent extension of existing products liability doctrine

An issue we're seeing at all levels of university is that many students are simply refusing to do *anything*. They aren't reading the syllabus, aren't following assignment guidelines, aren't engaging with material, ignoring deadlines. And this might seem like old news, but it truly has ramped up.

Short story Sunday: There once was a California newt heading down a leafy trail 1/??

periodic reminder of the existence of Atkinson Hyperlegible, a free font available from the Braille Institute designed to improve readability for people with low vision I use it in talks because it's pretty and also because, as an audience member, I am perpetually squinting at people's slides

Is it just me, or are LLMs ghosts made out of massive troves of written information that we can prod into taking on certain roles from their "memories" when talking to us?

Recently did a design project with a lot of dashboards, which made me realize I know nothing of data viz other than half-remembered lessons from AP stats. In the interest of improving myself, I'm going to liveread this from start to finish. www.oreilly.com/library/view...

Recently did a design project with a lot of dashboards, which made me realize I know nothing of data viz other than half-remembered lessons from AP stats. In the interest of improving myself, I'm going to liveread this from start to finish. www.oreilly.com/library/view...

Really cool article that explains the basics of ATproto and how it'll help create a less sucky social media! Very understandable to me as a nontechnical person. Thanks to @essentialrandom.bsky.social for the rec. overreacted.io/open-social/

Sammy Blue Jay taking off from the feeder this morning showing those amazing colours and designs.