Announcing, 50% sale of my app : AuraTracker:Habits & Goals, thru the month of June. AuraTracker is one productivity app that replaces the usual stack of habit tracker, reminders, goal planner, task manager, focus timer, and journal with one coherent workflow that helps you set your goals, plan your day, do the work, and reflect. No ads, no subscriptions, account setup needed. Purchase once and own it forever. More about the app : gitgudapps.com/auratracker iOS store link: apps.apple.com/us/app/habits…

And upgraded to Google AI Pro, for some reason, antigravity IDE doesn't show Argon , any idea why ?

Google omni 1.1 has improved a lot in video generation space, dialogues are more natural, facial expressions are better, it reads references more accurately and if you are detailed in the prompt, the results are much better. It's still not perfect, but so much better than what it was few months ago. I've been using it with Google Flow, thinking of upgrading from plus -> pro just for the 1000 monthly credits + 50 daily credits.

how on earth can posts like this which offer little to no value stack up so many views ? like what's the algo thinking recommending these posts to anyone - why on earth would it assume stuff like this is even remotely interesting or engaging ? Is it simply based on the no of followers the guy has or am i missing something ?

People who keep building these travel ai agents, don't understand a simple fact, a big "fun" part of the travel is in planning, of scouring the internet looking for places to visit, imagining yourself there, of booking hotels, checking out all the amenities comparing them, reading reviews, checking out discount offers, the high you get when you actually get a discount etc. Outsource all that to an AI agent and honestly, the joy of travel is 50% gone !

I'm not against open source, but having worked for 20 years in software dev, it feels really bad to see people's hardwork over the years, completely reduced to cheap clones so easily ! Wonder what's the future for software developers if a company as big as Adobe is powerless against IP theft: petapixel.com/2026/10/07/som…

Apparently i've hit 92% of my weekly limit 2 days before the time. Not sure if I should take it as a badge of honor that I'm working really hard or that I'm relying way too much on the AI agent and should scale back :)

How did Gemini lag so much behind other ai models ? Not just the llm, even their client app is so broken. Still doesn’t have basics like projects and context that’s there in gpt and even Claude and grok now !

Did @AnthropicAI reduced the quotas at their end or something ? I hit the 5 hourly limit and almost half way through my weekly limit with opus 5.5 (just 2 days of work), first time after launch ! Anyone else facing this ?

Learning of the day: Spent a while today fixing layout bugs (worst kind considering i'm a frontend noob) ! I check UI states by screenshotting the app with headless chrome. A short window looked broken, cards spilling past the bottom etc., so I measured element heights from inside the page to debug it, the window was 700px tall. the page said the viewport was 613. Turns out "--window-size" sets the browser window, not the viewport, your page gets less height than the number you pass. The irony was there was a real bug underneath, just not the one I was chasing - a grid row sized auto was growing to fit the sidebar instead of the window. grid-template-rows: minmax(0, 1fr) fixed it Then later in the day, an image crop made a perfectly fine header look scrolled off screen. The crop tool crops from the center by default. So yeah, two false alarms in one day, both from the screenshot pipeline, not the layout. Learning is: When a screenshot shows a bug, check the screenshot first. ps: I mostly use claude to build and debug issues, but it didn't catch this either. Claude treats the screenshot as the source of truth, ask it to fix a layout and it goes looking for css problems, it doesn't check whether the image itself is lying. Do let me know if you found similar issues when building and how your debugging pipeline handles it.

Grok Bot is really cool for running some background research when you are working on some coding work, wish claude would bring a similar bot feature soon. Also, looks like now they have added functionality of a primary bot that co-ordinates with your other bots and gets things done. I can't think of a use case for it yet, since my workflow isn't that complicated. But on paper, looks like another cool feature to have.

Claude sometimes just gets stuck without any reason like "thinking..." today it got stuck for like 7 mins for me. I stopped and restarted and it continued. I don't mind but the only issue is it consumes tokens throughout ! Any idea if it happens due to long context window or something else ?

the more you read about how to do something right, the more you feel its not for you. irony ?

Biggest superpower you can build for yourself is to train yourself to wake up at the same time (preferably before 8 AM) every day. It's so damn difficult esp without a 9-5 job !

If you're building macos app, here's a cheap trick for checking a Tauri (or any webview) app's layout without clicking through the real app: The frontend is just a web page. it only breaks in a browser because the native plugins (db, notifications) aren't there , so I stubbed those imports with fake modules in vite. alias: { '@tauri-apps/plugin-sql': './mocks/sql.ts' Then have the claude agent run: chrome --headless --screenshot --window-size=1440,700 http://localhost:5199 , Now my claude agent edits css, looks at the png, fixes any issues based on the test cases and I just review the end result :)

Some lessons I learnt while building macOS app using tauri : If your Tauri app asks for macOS permissions (calendar, contacts, camera etc), this can sometimes cause issues in dev. macOS requires a usage description in the app's Info.plist before it shows the permission prompt. Without a description, it kills the app instantly without any prompt / user friendly error message. `tauri build` puts your Info.plist inside the .app bundle, so release builds are fine. But `tauri dev` runs the bare binary, and a bare binary has no bundle around it to hold an Info.plist. The fix is to embed the plist into the binary itself, from build.rs: println!("cargo:rustc-link-arg-bins=-Wl,-sectcreate,__TEXT,__info_plist,Info.plist"); the linker copies the file into a section of the executable, and macOS reads it from there. dev and release now behave the same :)

Started witcher 3: remastered on steam, the base game doesn’t look as good as what ppl are claiming on social media (at least to me), the screenshot below is from hearts of stone dlc, substantially better !

What's your fav (preferred) layout while building ? I use a warp editor for most of my coding work. I have set it up on one monitor (with three panes : one vertical for claude code that takes half the screen, the other half split into two panes : one to monitor errors/logs and the other one to run shell commands on terminal). The iOS sims (in case of mobile app dev) or browser (in case of saas) runs on the other monitor. Both monitors are 27", anything above that and it causes me a significant neck pain !

So, icloud sync was not working for my app, it kept failing silently, reloading the old builds into my iOS sim. After some debugging, I found out that, apparently apple had sent a new dev agreement some weeks back and since I had not signed it, it just took away this capability ! Check this section under signing & capabilities if you are facing similar error. The text against the containers should be black , not red. If it’s red, sign in to your Apple account, sign the new dev agreement and then come back and refresh this. After the above steps, I did a force start of my apps on my sim as well.

So I just posted yesterday asking for what’s a good backup AI code gen tool to use along with Claude (when it runs out of tokens) and they did this. Awesome. Thanks @AnthropicAI !