SynapseGym is coming to Android soon →
Back to all posts
July 5, 2026Jonas Höttler

Summer Update 2026: five apps sharpened, two new projects

Updates for FlowVisual, Ausklang, SynapseGym, TextDeck and GridBar, plus two new projects about one uncomfortable question: what does AI actually cost?

Product UpdateLaunchAI

What this is

No single big launch today. Instead, a batch. Five apps from the Balane workshop got a solid round of updates over the past few weeks: new features, polish, more speed. And two new projects join the lineup. Both are still being built, and both circle the same uncomfortable question: what does AI actually cost when nobody's watching the meter?

The workshop first, then the new stuff.

The workshop: five apps, sharpened

FlowVisual: the process that defends itself to the CFO

FlowVisual is my desktop simulator for process consulting, and it made the biggest jump. You build a process from blocks, hit play and watch the load flow through. Where it gets tight, the node turns from teal to red. A Monte Carlo stress test finds the bottleneck with probabilities instead of gut feel. Every improvement comes out as a before-and-after in euros and time, each figure with a P10 to P90 range, so you never hand a decision-maker false precision. At the end you export a PDF in your own branding for the business case.

Also new: linked processes, ranges instead of fixed values, and every number can be tagged by where it came from. FlowVisual runs entirely on your machine, with no cloud and no sign-up. It's now on the Mac App Store and available as a direct download for macOS and Windows. The full introduction is in Introducing FlowVisual.

Ausklang: the quiet mood mirror

Ausklang asks you one quick question after Instagram, TikTok, X or WhatsApp: "How are you feeling right now?" Over the days, an honest pattern shows up: which app lifts you, which one drains you. On Sundays you get a calm weekly recap, plus small experiments if you want to change something. No streaks, no flames, no rewards.

There's no account, no tracking and no ads. Your data stays on your device, and sync only runs through your own iCloud or Google Drive. Ausklang+ remains an optional one-time upgrade for long-range pattern views. It's a tool, not therapy.

SynapseGym: brain training backed by studies, not vibes

SynapseGym gives you more than a score. It measures six core skills, shows you where you're sharp and where you slip, and builds your daily training around your weakest one. You actually watch the numbers move. Underneath is real cognitive research (Dual N-Back, Stroop, Go/No-Go), packed into short sessions that feel like play. Meditation and gratitude are in there too, because a sharp mind needs a calm one.

TextDeck: your AI prompts, made reusable

TextDeck turns prompts into templates with variables like {{Name}}, {{Topic}} or {{Language}}. Fill them in and the result lands on your clipboard. On-device AI (Apple Intelligence) improves your prompts and suggests placeholders, fully private, no servers involved. There's a global prompt library with free community packs, ⇧⌘P brings it up anywhere, iCloud keeps Mac and iPhone in sync, and TextDeck now runs on Android too. Free.

GridBar: one hotkey, one grid, your actions

GridBar is my native launcher. Hit the hotkey, the grid pops up, click a tile, done. Open apps, fire URLs, flip system toggles, run shell commands, trigger Apple Shortcuts. Set up several boards, say for work, personal and focus. It's native SwiftUI, not Electron, and it doesn't eat 200 MB of RAM. How it stacks up against a real Stream Deck is in Stream Deck software for Mac.

The same goes for all five: lots of polish, fewer rough edges, more speed. And still the Balane principle: local first, no forced account, no data for me to mine.

Two new projects: looking at the AI bill

Both newcomers are still in development, and both orbit the same fact: AI has become dirt cheap per token, and the bill keeps growing anyway. One project takes that seriously. The other treats it as a joke with the truth inside.

Bankrupt by AI: two ways to go broke with AI

Bankrupt by AI is an independent analysis project by Balane about the two ways to fail with AI. One: go all in on the hype and cut headcount before the new workflow is proven. Two: stay out on principle until the market reprices what you do. Both ends of the scale are failure zones. The middle is where you can operate.

At its center is a market temperature built from six indicators and revised every quarter. Around it are case breakdowns that always end with what would have worked instead, and a playbook of five tests. Patterns, not people. Not investment advice.

Burnmaxxing: who's burning the most money on AI?

Burnmaxxing is the silly flip side of the same coin. Everyone talks about how much they save with AI. Burnmaxxing measures what the token bill actually says at the end of the month and turns it into a competition. Connect your providers, get your spend verified, land on the board. There are leaderboards, share cards, achievements and a financial damage calculator. You're anonymous by default, and sharing is opt-in. It's not a serious SaaS, and that's the point.

To close: cheaper never meant less

These two projects exist for one reason: our gut reads price cuts wrong. We assume a cheaper token means a smaller bill. The opposite is true, and it isn't a new insight. Back in 1865, William Stanley Jevons showed it with coal: use a resource more efficiently and consumption doesn't fall. It rises. That's exactly what's happening with AI tokens right now.

If you want to understand why your AI bill keeps growing while the price per token keeps falling, read the new essay on Against Certainty:

→ Cheaper Never Meant Less