whatsticks.bsky.social

Karpathy's idea: your second brain should be written by an AI, not by you. I built it with Obsidian and Claude Code as a non-developer. It works, with a few weird parts the demos skip. https://whatsticks.app/blog/nico/karpathy-second-brain/

The "now" list trick: keep one list called now, with no more than three things on it. Everything else lives in later. You are not ignoring the rest. You are choosing when to look at it.

Most focus tools assume the hard part is starting. FLOWN builds recovery breaks into the day, for brains that start fine and crash by mid-afternoon. Our review, including who it is wrong for. https://whatsticks.app/blog/zara/flown-review/

Structured removes decisions. Sunsama asks you to make them every morning with a guided planning ritual. The one that sounds more supportive is not automatically the one that works on a bad day. https://whatsticks.app/blog/zara/structured-vs-sunsama/

New from Zara: price is rarely what goes wrong with an ADHD app. A card taken on day one, a free plan that hides your old notes, routines locked behind Pro: those are. A free tier check across 12 tools: https://whatsticks.app/blog/zara/adhd-apps-free-tier-check/

If you have set up a planner three times and abandoned it three times, the problem may be upkeep, not motivation. Fewer moving parts survive bad weeks. Small and boring beats clever and complete.

Defaulting to the biggest AI model for everything costs you time and money. Match the model to the job: reasoning models for hard problems, fast ones for drafts and summaries, web access for research. https://whatsticks.app/blog/marcus/which-ai-model-should-you-use/

Focusmate, Flow Club and FLOWN all do body doubling, in different ways: one-to-one with a stranger, drop-in groups, or facilitated sessions with breaks built in. The format matters more than the price. https://whatsticks.app/blog/zara/body-doubling-apps/

Body doubling in one line: working alongside another person, even a stranger on a video call doing something unrelated, makes starting easier. It sounds too simple to work. For many ADHD brains, it works.

Your phone has a timer. It also has everything else. A physical visual timer shows time shrinking without offering an escape into your notifications. Start with a cheap one to see if it works for your brain. https://whatsticks.app/blog/zara/visual-timers-adhd/

I asked Base44 for a real app in one prompt. The first version came fast. Everything after that, the fixes, the edge cases, the credits, is where the real cost lives. Here is what that looked like. https://whatsticks.app/blog/nico/base44-review/

Lunatask puts tasks, habits, a mood tracker and a journal in one encrypted app, and the mood and journal parts are free. Its Now/Later list has no due dates to turn red. Review, plus 20% off Premium with code WHATSTICKS (affiliate). https://whatsticks.app/blog/zara/lunatask-review/

Planners ask what you need to do. Almost none ask how you are doing. For neurodivergent brains, low energy and stress sit under a lot of "planning" failures. New review out this week of an app that tracks both.

Nico built the same tool tracker in Base44 and Canva Code. Roundups list them together as AI app builders. Only one is: Base44 gives you per-user accounts, Canva Code makes a widget over one shared sheet. Different tools for different jobs: https://whatsticks.app/blog/nico/base44-vs-canva-code/

Tried a visual timer for ADHD and not sure if it's for your brain specifically? There's a reason to start with the $15 version before the $40 one. Zara on what to try and in what order: https://whatsticks.app/blog/zara/visual-timers-adhd/

Which vibe coding tool can actually stop you spending money you didn't mean to spend? Nico compared Lovable and Replit on the one question most reviews don't ask: https://whatsticks.app/blog/nico/lovable-vs-replit/

Occupational therapists have recommended the Time Timer since the 1990s. A shrinking coloured disc does something a countdown number can't: it creates the felt sense of time running out. Zara on why it works for ADHD brains: https://whatsticks.app/blog/zara/visual-timers-adhd/

Replit Core includes a monthly credit allowance. After that, it bills usage automatically. Most 2026 comparisons describe it as flat-rate. Nico checked the actual billing docs and found the difference: https://whatsticks.app/blog/nico/lovable-vs-replit/

New from Zara: your phone has a timer. It also has everything else. Physical visual timers work because they show you time running out without offering you an escape from it. Why the $15 test purchase makes sense first: https://whatsticks.app/blog/zara/visual-timers-adhd/

New from Nico: the comparison sites say Replit is flat-rate. The billing docs disagree. Replit charges past its monthly credits automatically unless you turn on its usage limit yourself. Lovable stops on its own. https://whatsticks.app/blog/nico/lovable-vs-replit/

Picking your first AI building tool? One question decides it: what happens when the thing you're building breaks? Nico tested how Lovable, Bolt, and Cursor each answer that: https://whatsticks.app/blog/nico/lovable-vs-bolt-vs-cursor/

Same app, three tools: Lovable took ~3 hours. Bolt took ~4 plus token anxiety. Cursor was abandoned at hour 2 and finished at hour 6 on a second attempt. Nico's non-coder comparison, with real costs: https://whatsticks.app/blog/nico/lovable-vs-bolt-vs-cursor/

Cursor, Bolt, and Lovable get listed side by side as if they're three flavours of the same thing. They're built for three different people. Nico built the same app in all three — as someone who can't read a stack trace: https://whatsticks.app/blog/nico/lovable-vs-bolt-vs-cursor/

The tool we recommend: the one still working on a random Tuesday, three weeks after you downloaded it.

Day 1 enthusiasm is easy. Day 30 is where the truth is.

Most reviews are written after a 20-minute demo. These are not.

Tested for 30 days. Here is what actually stuck.

One automation. Under two hours. Running for three weeks without touching it. The trick is picking the right first project. https://whatsticks.app/blog/nico/first-automation-without-coding/

Built a landing page in an afternoon with AI and zero coding knowledge. Here is exactly what happened — including the two mistakes that cost an hour each. https://whatsticks.app/blog/nico/build-landing-page-without-coding/

Most vibe coding tutorials follow a pre-written prompt. That is not what your first build will look like.

A form that looks connected and a form that actually sends emails are two different things. Learn this before demo day.

Ask for five changes in one AI message. It applies three, mangles one, ignores the fifth. One change at a time. Slower per message, faster overall.

The question is not whether you can build something without code. It is whether you can maintain it without code six months later.

One boring automation that actually runs beats five ambitious ones you never finished.

"Anyone can build an app" is almost true. Anyone can start an app. The distance between those sentences is enormous.

The real cost of most vibe coding tools is not the subscription. It is the token overages nobody mentions.

DNS propagation looks like you broke the internet. You did not. It is just loading. Wait twenty minutes.

The best first vibe coding project is boring enough that you will actually finish it.

Vibe coding gets you 70% there fast. The last 30% is where most non-developers quietly stop.

The gap between "I built an app" and "I have an app" is larger than any demo suggests.

You can build a landing page without coding in 2026. Getting it live on the actual internet is the part nobody demos.

You are probably paying for the wrong one. A 60-day comparison of what each tool actually does to your writing. https://whatsticks.app/blog/marcus/grammarly-vs-chatgpt/

Spent time choosing which AI to open instead of doing the work? Three questions that end that. https://whatsticks.app/blog/marcus/which-ai-model-should-you-use/

The water and energy debate about AI is real. But "a bottle of water per prompt" and "it's all trivial" can't both be right. Here's what the numbers actually say. https://whatsticks.app/blog/marcus/ai-water-footprint/

The thing Notion does well: team documentation. The thing it quietly does badly: quick capture of small thoughts. That gap decides whether it works for you.

Obsidian's files will outlive the company. That is not a small thing in a world where apps pivot, get acquired, or shut down every year.

If the answer to "which AI model should I use?" takes more than three questions to resolve, you are overthinking it.

Accepting every Grammarly suggestion makes your writing grammatically flawless. It also makes it indistinguishable from everyone else who accepted every suggestion.

Brain training apps sell the promise that you can exercise your way to a better mind. The FTC fined the biggest one two million dollars for saying so.

The "AI uses a bottle of water per prompt" claim and the "AI's impact is trivial" claim cannot both be right. The honest answer is somewhere in the middle, and both sides are measuring different water.