What moved on the App Store: week of 24 August 2026

Every day we snapshot Apple's charts in the US, UK, Turkey, Germany, Japan, South Korea and China, across all 25 categories and three chart types. This is the Week of Aug 24, 2026 run: chart moves measured over a seven-day window, review evidence from each app's most recent sample, 294 scored ideas.

It is also the first run to carry China. That matters less for the ideas it adds than for the comparisons it makes possible: the same app, in the same week, priced and resented differently in seven storefronts. Most of what follows only becomes visible when you can put those seven side by side.

The biggest climbs

Chart movement is a two-point diff, so it needs two days of history and nothing else. No estimates, no modelling. These are the largest free-chart jumps in the window:

1. R-navi+86

   #87 → #1 · Sports · JP · free top 100

2. 一起作业（原一起小学学生）+83

   #97 → #14 · Education · CN · free top 100

3. 班级优化大师+82

   #93 → #11 · Education · CN · free top 100

4. Hatsune Miku official Mikunavi+80

   #94 → #14 · News · JP · free top 100

5. WeatherX: AI Weather Radar+75

   #89 → #14 · Weather · TR · free top 100

6. MapQuest GPS Navigation & Maps+73

   #75 → #2 · Navigation · US · free top 100

Two of the six are Chinese school apps, and the last week of August is when Chinese term dates land. A climb like that is a calendar, not a product decision, and it will unwind on its own. The one worth a second look is MapQuest, because US Navigation is not a category that usually admits anyone.

What it costs to enter a top 100

We model an entry wall for every category and storefront: roughly how many ratings an app needs before it can hold a top 100 slot. Pick a storefront below; these are the cheapest doors each one had open this run.

Same store, same week, and the widest gap in the run is 9,685,904 ratings for Entertainment in China against 8 for Navigation in South Korea. Picking the storefront is a bigger decision than picking the feature, and almost nobody treats it as one.

One pattern survives every storefront in that picker: Developer Tools is the cheapest door in Japan, Germany and the UK, and close to it in Turkey and South Korea. A category whose users are the same people who could build the competitor turns out to be the least defended one on the store, in five countries at once.

The same category is priced three different ways:

• Navigation: about 36,785 ratings in Japan, about 8 in South Korea, a 4598x gap.
• Magazines & Newspapers: about 288,969 ratings in the US, about 134 in South Korea, a 2156x gap.
• Reference: about 298,841 ratings in the US, about 223 in South Korea, a 1340x gap.

Charting, and hated

A chart position says an app has distribution. It says nothing about whether the people inside it are happy, and the gap between those two signals is where the openings are. Counts are how many of the sampled reviewers said the same thing.

Both of these are top-grossing apps whose recent reviewers are almost uniformly unhappy, and in both cases the anger has a single named cause rather than a general drift. That is the useful shape. A product people have stopped liking is hard to attack; a product people still want, delivered badly since a specific change, is not.

What scored highest

Every idea is scored 0–100 on demand, winnability, monetization, feasibility and momentum, with a penalty for picking a fight with a giant. The scoring engine is public, weights included, so you can disagree with a number rather than take it on faith.

• ParcelProof

  Shopping · US · week to build

  94/100

  Takes from Package tracker: Parcels Hub · Quick win

• Tansu

  Shopping · JP · week to build

  87/100

  Takes from XZ(Closet) Fashion Outfits · Quick win

• 旅ポケット

  Travel · JP · week to build

  87/100

  Takes from まっぷるリンク - 自分にぴったりな旅が見つかる・作れる · Quick win

What this run cost

Building in public means the bill too. This run took 15,774,046 model tokens across 25 categories, 86 agents and 15.2 hours of agent time, about 53,653 tokens per scored idea. 6% of those tokens went on retries, which is the number worth watching: it is what tells me a stage is quietly half-working.

Every token of this run's retry figure traces to the machine it ran on rather than to the work: a laptop that went to sleep twice and an endpoint that stopped answering for a while. That distinction is the reason the number is published at all. Rework caused by a bad prompt is a quality problem worth fixing; rework caused by a closed lid is just a closed lid, and conflating the two hides the first inside the second.

Check it yourself

Weekly runs go to subscribers, but a complete monthly run is published free and in full: all 25 categories, every score and the evidence behind it. Older monthly runs land in the archive as they age out. If you only want one number, the keyword checker is free and needs no account.