Chapa
Go, mostly. Card-to-card, gift cards, and a webhook dispatcher that has to be right every single time. Money is unforgiving code. Nobody notices it working.
PaymentsGo backends, webhooks, the boring reliable kind
Researchhuman state, heart signals, stubborn sensors
AIWAR
Anwar Misbah — payments engineer, AI researcher, and the person testing whatever tool promises to make all of that faster.
Anime, sort ofa youtube channel with no face in it
AgencyAI support that answers before I wake up
The newsletter · one email a week
AI, Actually Useful
Five tools I actually used this week, each with a verdict — kept, dropped, or watching. One deep cut with the config that made it work. One hyped thing to skip.
where the verdicts come from
Everything gets tested on real work
A tool recommendation is only worth what it was tested against. These are the four places anything I write about has to survive first.
- Production systems that cannot fail quietlyPayments backends, where a tool either survives real traffic and real money or it does not get used twice.
- Applied AI researchSignal processing and model work against messy sensor data — the setting where most "AI-powered" tools quietly fall apart.
- A content pipeline that runs unattendedScripting, generation and rendering that has to work without a human watching. Great filter for anything that needs babysitting.
- Client work with a deadlineAgency builds, where the question is never "is this interesting" but "does this ship on time".
Latest issue — #2: The cheapest AI upgrade is writing down what you actually want 22 Aug 2026
pinned to the wall
Four things at once
Two jobs and two of my own. People keep asking how they connect. They connect more than they should, which is the red string.
Synheart
Teaching machines to read human state from a heartbeat. Half research, half wrestling sensors into telling the truth. The wrestling is most of it.
Anipreneur
Self-improvement ideas explained through anime characters. No face, no camera. The whole pipeline is code, which is the only reason I can run it alone.
Inlinkai
Support automation for B2B SaaS. Started as "I can build that", turned into an actual offer. Currently the most business-shaped of everything here.
longer than a telegram post
Writing
Written after the work, not instead of it. All posts →
The only AI benchmark that matters is your own work
Public leaderboards measure something real. It is rarely the thing you are about to use the tool for. A cheaper test: hand it three problems you have already solved.
Read it →Pick the stack you can debug at 2am
Running several projects alone changes what a good technical decision looks like. Novelty stops being interesting and starts being a tax you pay at the worst possible moment.
Read it →taped up next to it
And the rest of it
Smaller things. Some earn money, some earned a lesson, all of them shipped.
aiwar.dev
A bot that reads AI news so I do not have to, summarises it, and posts it. Named before I had a website.
fact.et / forbes.et
Two Ethiopian business publications. Editorial systems, not just a blog with a nice header.
a second brain
A wiki that documents itself. Every session writes what changed. Slightly cursed, extremely useful.
Saved Messages, out loud
I spent years sending every half-thought to Saved Messages. Thousands of messages, zero replies, very healthy. Now they go on a Telegram channel instead, four or five times a day, mostly about whatever broke that morning. The newsletter is the weekly edit of it — the part that survived the week.
Read @aiwar_dev