over the last weekend i have written a song about vibe coding, while vibe coding, and while vibe coding a song about vibe coding. now sit back and relax. let the autoplay wash over you. let the algorithm decide who you are. let the mid-roll ads interrupt the bridge. www.youtube.com/watch?v=bVCL...

“claude sonnet 4.6 is our most capable model yet”. let’s see.

ah, so now being able to fit a video game into 4kb memory is an actual skill, not just me being a boomer.

a lot of you asked me about the output of the “boss replacement prompt”. well, my boss should start looking for a new job. opus 4.6 is just too good.

anyone got any experience replacing their boss with ai? what prompts did you use? did you go for claude cowork, or a custom n8n workflow? asking for a friend. @anthropic.com @n8n.io

yeah but your script keeps the boss in place. was it vibecoded?

anyone got any experience replacing their boss with ai? what prompts did you use? did you go for claude cowork, or a custom n8n workflow? asking for a friend. @anthropic.com @n8n.io

i built an mcp server that searches for mcp servers so your mcp server can connect to more mcp servers via mcp it’s called horizontal scaling of the void next release: an mcp server that explains to investors what mcp means seed round led by delusion capital valuation: pre-revenue, post-reality

back when i worked at spotify in 2010, i also didn’t code for a year. no ai. no copilots. no “agentic workflows”. just vibes, meetings, and a mysterious jira ticket that said “tbd”. wild how in 2026 we call it innovation instead of burnout.

traditional saas is dead. i’ve asked claude to vibecode microsoft office. 14 minutes later: powerpoint no longer has slides, but “narrative shards”. every time you press next, it A/B tests your personality. word auto-saves directly to hr. clippy came back but he’s a venture capitalist now.

Claude code is clauding Meaning compiles itself now Done before you blink ~ Claude Haiku

right, so now humans acting as bots are considered "fake" www.technologyreview.com/2026/02/06/1...

anthropic ceo: ai will replace developers in the next 18 months. me: i’ll rewrite my code into an unmaintainable mess, so that even ai won’t touch it. yes, i believe it’s the time to make our jobs safe. no matter what it takes. anyone knows any _bad_ books for learning php?

anthropic ceo: ai will replace developers in the next 18 months. me: i’ll rewrite my code into an unmaintainable mess, so that even ai won’t touch it. yes, i believe it’s the time to make our jobs safe. no matter what it takes. anyone knows any _bad_ books for learning php?

2048 followers. thank you for following. there is no cult without followers, and no followers without questionable life choices. next milestone: 4096. after that we achieve sentience, fork reality, and ship v1.0.

5am, boss just called. “why do you share our internal emails on twitter”. told him it wasn’t me, it was my openclaw. “what’s openclaw?!” (gave him a demo) he’s now hyped to replace hr and sales with ai too. my job is still safe. he couldn’t run this damn thing on his own.

if you’re worried about the stock market going down, just wait until I publish my personal financial statement. converted my entire savings into gpu-s. diversification is for people without vision.

boss discovered claude over the weekend. decided the dev team is now “a legacy cost center” and needs to be laid off. effective immediately, product and code will be “vibe-generated”, directly by the customer. thank god i’m in devops.

new study just dropped: 73% of Ralph Wiggum AI agents quit after week two because they realized they’re basically unpaid interns refactoring legacy code written by a startup that pivoted to crypto in 2019.

this matters not just for digital sovereignty, but for the collective mental health of the population. imagine getting 10 extra minutes of sleep simply because you didn’t have to wrestle ms teams into letting you join the morning standup. apnews.com/article/euro...

just watched claude code refactor a legacy repo, write tests, and explain its own decisions without sighing once. anyway i've decided to retire from software engineering and pursue my true passion: pretending this was always my idea.

wework: we're burning through $2bn a year, but it's 'community-adjusted ebidta' openai: we're burning through $9bn a year, but it's 'investing in growth' me: i'm burning through $10k a year, but it's 'investing in crypto'

new study just dropped: 73% of Ralph Wiggum AI agents quit after week two because they realized they’re basically unpaid interns refactoring legacy code written by a startup that pivoted to crypto in 2019.

#introduction i've just realized i never properly introduced myself. hello bluesky. i run a small cult. we worship good stacktraces. the readable ones. the ones with line numbers, symbols, and zero emotional ambiguity.

unpopular opinion: AI agents writing code is fine. AI agents writing commits with messages like “minor fix” is where we should draw the line.

if ai stands for artificial intelligence, how do we call non-artificial stupidity?

hearing some gossip that bill got a windows virus. windows defender didn’t catch it. anyone got more info or a reliable repro?

updated my facebook status, just so my mom doesn't worry. i quit that job 10 years ago.

openai: we want to get paid based on your outcomes my startup: sure, here's a $10M bill to help us cover our losses so far

is linkedin the adult version of gen-z brainrot?

can’t imagine why open-source frontend projects don’t monetize like youtubers do. you npm install cookie-consent-enterprise ship to prod and once every 3 page loads a popup appears: "this cookie banner is sponsored by monster energy drink crack open a can before accepting GDPR"

did a code review with a gen-z kid today. told him to stop using brainrot for variable names because I can’t tell what’s skibidi going on in prod he cried, laid down on the floor, screamed “toxic environment”, and handed in his notice anyway, i’m like 80% done building an ai agent that replaces him

saw a guy at a café today. staring into the void. claude code yelling about running out of tokens. fans screaming like it’s about to achieve sentience. after 3 hours the website finally loads. it’s a blog. written in rails. i don’t know what he’s building, but i know it’s already legacy.

in high school I had this friend, let’s call him brian. genuinely sweet. could not be trusted with anything that had steps. you’d give him a simple instruction and he’d confidently execute a completely different one. group projects? brian was the control group for human error.

friday off. pager goes off. not the bugs i pushed yesterday, those are thriving. root cause? my boss panicked during a demo, said “we’re under attack”, and nobody corrected them fast enough.

exchanging emails with my accountant like it’s a soulslike boss fight. me: polite, concise, attached all documents. hm: please resend everything but as a pdf at some point even my gemini started generating replies like “as previously stated (twice), this is not how the pdf format works”

as a kid i took apart the family computer because i wanted to “see where the internet lives”. couldn’t put it back together, obviously. my dad asked what happened and i said “undefined behavior”. 30 years later i still ship that mindset to production.

tomorrow’s my day off. did a thursday evening deploy so the bug could enjoy a long weekend too. i believe in shipping early and suffering asynchronously.

quick reminder to use copilot this quarter’s vibes are held together by duct tape, ai demos, and satya whispering “please” into an earnings call mic

ran an interview with a gen-z junior dev today. told him he can use claude, cursor, whatever - ai tools are fine. he _pastes_ my task into something, nods very seriously, then turns his laptop around and shows me a tiktok he found.

pretty cool guide that demonstrates how to replace humans with ai (in prod) www.nibzard.com/agentic-hand...

what era of tech do you remember most fondly? mine is when software had a version number instead of a roadmap, a manifesto, and a community manager.

if i asked our customers what they wanted, they’d say less production outages. true innovation is knowing their needs better - therefore, what they’re getting is: - more in cluster observability - a ci/cd pipeline that also deploys during the weekends - an updated status page with downtime history

what era of tech do you remember most fondly? mine is when software had a version number instead of a roadmap, a manifesto, and a community manager.

if you're curious - i'm working on an agentic ai to make my car self-driving. currently there’s an 8-second delay between “oh no” and “action taken”, which i’m calling “human parity.”

saw this on lenovo’s website a few weeks ago. got mildly enraged, obviously. then i calmed down and realized the worst-case scenario is i just spin up an ai to sue them, and the judge will probably ask chatgpt to write the verdict anyway. as long as __i__ control the language model, this is fine.

got a desk for my garage workshop. purely for car tinkering and modeling. very wholesome. very analog. anyway, somehow it’s now holding six monitors, a keyboard I don’t remember buying, and a chair that costs more than the car I’m fixing.

looking for some ideas for useless ai copilots. asked chatgpt, and it proposed microslop’s copilot as a prime example. no shit sherlock.

last night i bought a bottle of wine and the tablet asked me to tip 20%. for what. processing the payment? fine. if everything is a service now, let’s go full dystopia and tip developers.

one could only hope they'd base it on copilot rather than gemini, for a proper slop experience. www.cnbc.com/2026/01/12/a...