DevDiary

Matte Tactile Mini-App Trackpad Hard to pull off, but it looks kinda cool. Make it optional(easily turn it off and make it look like normal trackpad with one button click)

teapawt.net/JamesMalsawm/status/21… Me rn: Claude: 20$ OpenCode: DeepSeek V4.1 Free Codex: 20$ + (Stole my sister's subscription 20$) Cursor: ~2$(New X subscription) Antigravity: Jio subscription (Mom's phone) Anyone else running a cheap setup like me?

Yes bro, heck we are doing the same thing as managers😭😭 "Fix code" "Do fast" "Production error, fix it, make no mistakes" "Is this feasible, make it work" "Fix UI, it's shit"

Many people are saying "Cursor is just for editing" - in that case why not use Zed or VS Code? Why do you guys use cursor? I'm super curious

teapawt.net/yihui_indie/status/211… Yihui launched Skillry one month ago with 400+ active paid subscriptions and $12,037 revenue in 30 days, crossing the monthly 10k USD mark!! TLDR; Before writing a single line he forces two answers: (1) What’s the actual business model that already works? (2) How do the first paying customers show up, and can I copy that? For Skillry the answer was “sell aesthetic templates/skills that AI coding tools can call directly, via subscription.” not another skill directory. AI-generated templates have that plastic feel and developers will pay for taste. Same old business as Tailwind UI ($299 one-time, $2M+ in 5 months), Magic UI / Aceternity UI (lifetime deals + pure X demos doing $60–100k/month), and 21st .dev (MCP into Codex/Claude Code). He used BizPeel to pull real last-30-day numbers on 7 peers and the pattern was brutal: > old one-time template plays (ShipFast still looks legendary at $1.27M cumulative but only $1.7k last 30 days; God of Prompt MRR dropping and listed for sale) are sunset > the ones still printing are “subscription + feed it straight into the AI tool” (Sleek $16k+/30d, TypeUI growing, Skillry itself) > focused deliverables crush generic ones. TypeUI (generic components) did ~$3.4k. Skillry (complete aesthetic pages, decks, promo animations you can actually ship) did $12k. Low Content AI just targets Amazon KDP sellers and clears $15k/30d with almost no broad SEO The main unlock is this: Reverse engineer a product, see if you can agentify it (make it available for codex/claude code). Usually what most of us(including me) do is build and then pray. Highly recommend to read the original article.

teapawt.net/GoogleDeepMind/status/… Google will win the AI race (atleast in AI video generation, ads and games)

Bro disappeared like he never existed

The biggest unlock for junior engineers: a database isn’t a black box. It’s a system with its own network, cache, and core logic. Postgres: Network → TCP, process-per-connection, pooling Cache → shared_buffers, OS page cache, thrashing Core → B-tree inserts, page splits, WAL

GPT-6.1 Terra just launched. Main pitch is fast computer use at lower cost. It can see what’s on screen and take actions across apps. Still too early for real tests, but the interesting part is the everyday stuff that used to need a human clicking around. Filling forms, moving data between tools, basic research then acting on it, editing things inside actual apps. If it holds up, a lot of the small repetitive work that eats time could finally get cheaper and less fragile. "But Opus can do it better" - yes ofc, but Opus isn't cheap. Low end models are enterprise favourites for a reason teapawt.net/OpenAI/status/21147807…

End of an Era, #Bleach was spectacular. Few hiccups at 48 and 49, but my nostalgic side of the brain stops me from overthinking. Thank you Kubo Sensei. Off to CFYOW and Hell arc