AI coding for 38h straight 😵💫 I have little faith anything is gonna work 🤣 what's your bet?
Dan
@d4m1n.bsky.social
Potentially not an award winning dev person. Likes to design UIs, but also likes pizza🍕 Shipped: 📄 pageai.pro 🎙️ morningmakershow.com shipixen.com / imgxai.com / hunted.space / crontap.com / clobbr.app / saventify.com more on: mindrudan.com / x.com/d4m1n
some days I dream it's only me and a handful of AI cowboys having this kind of grip on agents. but deep down... I know everyone has the same. it's kind of insane to think that we went from gpt-3.5-turbo 👉 not being able to output correct syntax to: *checks notes*
yeah I just lost a customer here. I would have never discovered and fixed this bug. there is no chance. zero. I f***** love Sentry so much
tbh I never saw one of these autonomous AI content tools result in sustained growth not the blog/SEO ones not the marketing/social automation ones not the cold outreach ones not the video gen ones atm they are all ways to sell hope to people that don't want to put in the work.
how good is GitHub Copilot in vscode as a harness anyway? I *NEVER* see benchmarks on that, but see Cursor, Claude Code & Codex benches every other week. GH Copilot have been in the game the longest, so it must be good?
if I could build an email client that filters AI spam I'd be a trillionaire
I still don't understand why CLIs became the defacto Agentic UI what's wrong with images and nicely formatted text?! IMO Cursor has a great DX 🧑🍳👌
Grok 4.5: $2 in / $6 out GPT 5.6 : $5 in / $30 out Opus4.8: $5 in / $25 out Fable 5: $10 in / $50 out someone's either ripping us off or making a huge loss on us
if you were to build the dream LLM benchmarking tests, what would they be? I'm thinking: - landing page - dashboard - game - 3d viz what else?
lmao GPT-5.6 Sol completely ignored my prompt. tl;dr the user's an idiot I know what's best what a strange model 😃
Cursor Cloud Agent setup is go goated hot take but their cloud DX is the best in the biz I even got a video of the running app?!
every. single. damn. time review with different models folks, it's worth it
testing Fable 5 vs Opus 4.8 at landing page design can you tell which is which? :D
everyone’s gonna end up shipping WASM to canvas by 2027 to avoid being cloned
yeah y'all talk smack about Anthropic say you'd never work there but if they presented you this offer... would you have it in you to say no? be honest. I'd decline.
People say DeepSWE is the only benchmark that still matters. So here are the results. Claude Fable 5 was a leap I only shared this with a few people, but it felt so much more capable in doing end to end complex flows. Perhaps it crossed a certain threshold, but you notice it easily. Not a fluke.
Opus thinking like an absolute *chad* for 9 min straight. have you also experienced this a lot?
i just keep thinking about it but my mind is blank how did Shadcn *know this*
loops are just tokenmaxxing on the company's LLM budget. if you're not doing multi-model slices (MMS) you are ngmi
PSA: you will pay the full cost of cache writes if you do do this 💀 DO NOT switch models late into the conversation, spin up subagents instead. I often spin up a gpt-5.5 review agent :)
not something on my 2026 bingo card Cursor has been my daily driver for the past 2 months IMO better in every single way vs Claude Code - harness is better - subagent & cloud agents better - planning waaay better I am *absolutely dreading* to open Claude Code CLI these days.
everyone shipping AI updates weekly redefining industries so Apple had to push their limits too
dude I refactored crontap.com pricing with Fable 5 it completely over engineered it like hell 😂 I love it haha
How to rank higher in LLMs according to Posthog one is not like the other 😂
