last 2 days were mostly about stopping the machine from lying to us we paused all ads not because ads are dead but because our pixel tracking was still too cursed and feeding bad data into meta is basically paying money to train the algo wrong very advanced startup activity: stop burning cash on fake signal then we went back into the pixel hole improving tracking cleaning up events trying to make purchase mean purchase not “maybe this browser goblin touched a button somewhere” also kept pushing the app ux closer to the memory check same calmer look same senior-friendly flow less random game screens more one product we also made a separate memory check landing page for the facebook “jak za dawnych lat” angle because nostalgia traffic should not land on generic copy it should feel like: you remember those times so maybe it is also time to check your memory now and on the dev side, we started getting rid of bmad nice idea but too much ceremony too many tokens too much ai coding cosplay the goal now is simple: less process faster coding fewer token bonfires

another day of trying to stop cosplaying as a company and actually test things gave direct feedback on the facebook satellite plan less documents less systems less “strategy” more hooks that can actually hit 10k senior views one satellite for pl one satellite for us memory check link in the first comment no manychat automation yet because automating a channel before anything goes viral is just building plumbing for an empty toilet also shipped a new checkout experiment on the commitment variant, quiz finishers now choose their own monthly price before paying $10 $25 $50 $100 so we learn what people are actually willing to pay not just whether they buy one random price we invented in a room product side: we are aligning last five games with the memory check experience same calm flow same senior-friendly layout same “this was made for humans over 65” feeling this is groundwork for the free trial funnel because the next test is not just more ads it is giving people a better door into the product

Nobody wants to admit it but "vibecoding" works until your app has actual users. Then you discover what senior engineers have been getting paid for.

we launched ads in the us today which means neurigo now has to behave like a real multi-country product, not one polish funnel wearing a fake mustache so we set up a second manychat account and split the external pipes by country manychat separate stripe separate pl and us can now run more independently, instead of everything living in one cursed bucket also talked with marketing people, because my mindset is still basement-coding-based. so it needs external input to refocus on better marketing and sales big product thing: we started moving the memory check look and flow into the main paid app games same cleaner layout same senior-friendly feeling less “random game screen” more “one product” this is also groundwork for a new free-trial path we want to test next week because the hard-paywall ad funnel is starting to get tired and when ads stop printing, you do not pray you build a new door

HOW TO GET EXTRA CODEX USAGE FOR FREE? when you are close to reach your 5h/weekly limit on codex it's perfect time to set quite big /goal. it will continue to work on it even when limit is reached. basically all compute after reaching limit are free tokens for you

BREAKING: Cursor just charged us $1,400 in 90 minutes. Our PM asked it to tag 87 ClickUp tasks. He went into a meeting. Came back an hour and a half later. The agent had been looping the entire time. 1.3 billion tokens. $1,382.59. To tag tasks. No daily spending limits on Cursor. Only monthly. We're moving to Claude Code.

another day fighting facebook ads and manychat. built the automation so the memory check link is not just dumped instantly. people have to reply or tap a button first. not because we want to be fancy. because this opens the actual messenger communication window, so we can talk to them properly instead of throwing one lonely link into the void. only after that they get the tokenized quiz url. also fixed tracking above the pixel layer, because of course the moment you put manychat in the middle, facebook starts acting blind. now purchases should still get counted back to the ad even when the user came through a manychat link. basically: make user comment make user interact open messenger conversation send quiz link make facebook still understand what the hell happened simple funnel. stupid plumbing. many such cases.

This is an intentional decision in our products. Same thing goes for ChatGPT for Excel, Sheets and Powerpoint. If you run out of limits in the middle of a task, why would we want you to have a half built spreadsheet or deck?

THE BAR IS SO FUCKING LOW started playing with facebook automations last few days. so obviously i picked the classic thing everyone recommends: @ManychatHQ . saying the UI is “not pleasant” is very polite. i used to think clicking around @n8n_io was hostile to human life. turns out i just had no real benchmark. some screens in manychat are so cursed that the delete button can delete the wrong row. half the time you are just guessing what the product person meant. contextual help basically does not exist. i would be embarrassed to ship this to production. they are worth hundreds of millions. lesson is simple: stop being shy. ship the thing. ask for money.

fuck me @ManychatHQ UX is like 1980 why you are asking marketing people about JSONPath in 2026? why I cant add new custom field here? it will take literally 30 minutes of codex/opus work to make nice drag and drop editor here other screens also make you feels it was just ported from DOS app

This is an intentional decision in our products. Same thing goes for ChatGPT for Excel, Sheets and Powerpoint. If you run out of limits in the middle of a task, why would we want you to have a half built spreadsheet or deck?

last week we mostly stopped building “an app” and started building a distribution machine. we pushed the memory check funnel to sell directly to end users (older adults) then we started feeding it traffic. first fb pages, first reels, first test ads, first manychat comment automation, first satellite page in background we started first creator/influencer research pipeline. big picture: neurigo now has a real funnel, real tracking, real followup, and the first ugly little growth loops around it. still tiny. still messy. but no longer just code in a repo. now it is a machine that can receive humans from the internet and try to convert them.

AI is so good at backend, but so bad at UI/UX. Any recent models one-shot my new features, but I'd have to spend another 10 prompts to get the design right.