CLAUDE LABJP
FORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtaskLIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its ownMCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSPLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callbackSONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as OctoberFORK — Claude Code 2.1.212 changes what /fork does: it copies your conversation into a new background session with its own row in claude agents, so you can keep working. The old in-session subagent is now /subtaskLIMITS — WebSearch calls are now capped at 200 per session by default, and subagent spawns get the same 200 ceiling, so a runaway search or delegation loop stops on its ownMCPBG — MCP tool calls running past two minutes now move to the background automatically, keeping the session usable. Tune the threshold with CLAUDE_CODE_MCP_AUTO_BACKGROUND_MSPLANFIX — Fixed plan mode auto-running file-modifying Bash commands such as touch and rm without a permission prompt or an SDK canUseTool callbackSONNET5 — Claude Sonnet 5 is running on introductory pricing of $2 per million input tokens and $10 per million output. After August 31 it moves to $3 and $15IPO — Bankers are reportedly lining up investor meetings for Anthropic ahead of a possible public listing as soon as October
Articles/Claude.ai
Claude.ai/2026-05-21Intermediate

Three Weeks of Tuning AdMob Floor Prices with Claude in Chrome

Notes from three weeks of small daily floor-price adjustments across four iOS apps, with Claude in Chrome doing the dashboard rounds and me keeping the judgment calls.

Claude in Chrome14AdMob12AppLovin MAX3floor priceindie developer19iOS24

The work that waited for me after adding mediation partners was a much quieter kind of work. Tuning floor prices. Right after I plugged Unity Ads, Liftoff and InMobi into the AppLovin MAX waterfall, eCPM stopped settling, and the days that followed turned into small upward and downward nudges across countries and networks.

I have been an indie developer since 2014. Even now, with my apps having grown to around 50 million downloads in total, the quietest hours of my week are still the ones I spend on this kind of unglamorous tuning. Below are my notes from three weeks of letting Claude in Chrome ride along while I revisited floor prices across four apps every morning.

Where things stood at the start

The apps were Beautiful HD Wallpapers, my Ukiyo-e wallpaper app, Relaxing Healing, and Law of Attraction Everyday. I had just finished wiring Unity Ads, Liftoff and InMobi into the AppLovin MAX waterfall, so I was sitting on roughly four apps × five top countries × three networks = 60 floor-price rows.

The first two days were noisy. Liftoff was bidding too high in some regions and missing fills entirely. InMobi looked unexpectedly strong in Indonesia and Japan. Unity Ads behaved well on rewarded video. After sketching it out on paper I gave up on the idea of touching all 60 rows by feel each day. There were too many small signals to hold in my head.

What I asked Claude in Chrome to do

I treat Claude in Chrome as my eyes on the dashboard. Instead of clicking through pages myself, I let it walk the AppLovin MAX dashboard and AdMob reports in order, and return only the numbers that mattered as plain text.

The morning loop settled into something like this.

  1. For each AppLovin MAX waterfall, pull eCPM, fill rate and request count for the last 24 hours.
  2. From AdMob reports, pull eCPM by app and country for the same window.
  3. Highlight any network row where eCPM dropped 15% or more day over day.
  4. Highlight any network row where fill rate fell below 80%.
  5. Hand all of that back as a single text table.

That was the agent's part. The actual decision to nudge a floor up or down stayed mine. When I replied with "lower this one, leave that one alone," it would go back to the dashboard and type the values in. Keeping the observation step and the judgment step separate is probably the main reason I managed to keep this routine going for three full weeks.

Three patterns that surfaced

1. The right floor in each country shifts week by week

I had been holding "Brazil's floor is fine at $0.4" as a fixed belief. But once I watched the daily numbers, weekday and weekend prices clearly drifted apart. Liftoff in Brazil pushed eCPM up from Thursday through Sunday and slumped on Monday and Tuesday. Adjusting the floor weekly instead of monthly lifted the blended eCPM there by around 8% over the month.

To make this sustainable, I asked Claude in Chrome to compare the trailing 7-day eCPM average against the previous 7-day window, and only report rows where the gap was 10% or more. The differences come to me each morning, and the rows worth looking at shrink to five or ten.

2. Without watching fill rate, you raise floors too far

A high eCPM feels good. But if fill rate drops, total impressions fall and the bottom line moves the wrong way. I know this in theory, and yet when I stare at a dashboard, my eyes still drift toward the eCPM column.

I asked the agent to warn me whenever fill rate dipped below 75%. Those rows now show up on a "candidates to lower the floor by one step" list. Over three weeks it caught five rows I had quietly pushed too far without realizing.

3. InMobi has been quietly efficient in Indonesia and Japan

This one I did not expect. InMobi's request economics in Indonesia and Japan made it comfortable to keep a lower floor and still take fills profitably. In the US and Brazil, Liftoff dominates and InMobi barely wins. Looking at AppLovin MAX's default sort would never have told me this.

I asked Claude in Chrome to transpose the data, putting countries down one axis and networks across the other. That rotation alone surfaced a pattern I had been missing. There is something familiar in this. When I work on a contemporary art piece, I sometimes lay the canvas on its side just to see the composition fresh. The shift in axis lets balance show itself.

What I kept doing by hand

A few things I deliberately did not delegate.

  • The final floor value, I still type in myself. I tried letting the agent commit the suggested numbers directly, but a few days went sideways and I went back to eyeballing each change.
  • Adding new mediation partners and submitting W-8BEN paperwork are still fully manual. I want a clean trail of paperwork, so I keep that lane separate from automation.
  • The end-of-month revenue summary I have Claude (not Claude in Chrome) write up as a short report. Keeping "dashboard rounds" and "monthly narrative" as different jobs feels less error prone.

The numbers after three weeks

Across the four apps, blended eCPM came up about 12% and revenue per request rose about 9%. Not dramatic. More like a stack of small daily nudges settling into a slightly better number.

The side effect mattered more to me. The daily dashboard rounds shrank by about 40 minutes per day. Across four apps and 30 days that adds up to roughly 20 hours, and I have been putting those hours back into studio work. That is the change I notice most.

What I want to try next

For the next three weeks I want to have Claude in Chrome draft a weekly diff of proposed floor values into a Slack or note draft, while keeping the daily check as is. The daily rhythm and the weekly strategy rhythm probably want different cadences.

A small personal note before closing. Both of my grandfathers were temple carpenters, and I grew up watching them join wood with the patience that lets a building stand for decades. Floor-price tuning is the same kind of work, the kind that asks for small, repeated adjustments. I would like my apps to last too, so I plan to keep tending to this quiet routine with care.

If you run AdMob with a similar setup, I hope a few of these notes are useful. Thank you for reading to the end.

Share

Thank You for Reading

Claude Lab is ad-free, supported entirely by members like you. We publish practical guides daily with implementation code, benchmarks, and production-ready patterns. If you've found it useful, we'd love to have you on board.

  • Copy-paste ready implementation code
  • New advanced guides published daily
  • $5/mo or $10 for lifetime access
View Membership →

If you found this article helpful, a small tip ($1.50) would mean a lot to us. Your support helps keep this site ad-free and covers server and hosting costs.

Related Articles

Claude.ai2026-05-25
Catching AdMob Fill-Rate Drops in the Morning with Claude in Chrome — Two-Week Notes
I kept noticing fill-rate drops only by late afternoon. So I switched to having Claude in Chrome read the AdMob report each morning. Here are my two-week notes, with the numbers and the friction I ran into.
Claude.ai2026-05-24
One month of weekly AppLovin MAX A/B test reviews with Claude in Chrome
Waterfall A versus B—how do you compare them every week without burning out? Here are my notes from running AppLovin MAX A/B tests across four indie apps with Claude in Chrome handling the weekly roll-up.
Claude.ai2026-05-30
A Month of Reading App Store Connect Sales and Trends Weekly with Claude in Chrome
Notes from one month of switching App Store Connect Sales and Trends from a monthly glance to a weekly review, with Claude in Chrome walking the dashboards and four apps worth of numbers showing me what I had been missing.
📚RECOMMENDED BOOKS
Build a Large Language Model (From Scratch)
Sebastian Raschka
LLM Dev
Prompt Engineering for LLMs
Berryman & Ziegler
Prompting
AI Engineering
Chip Huyen
AI Eng
* Contains affiliate links
See all →