Dan Martuszewski

Founder @ Agents Mountain
AI agents for real businesses - not demos.
15 yrs shipping SaaS. Now building agent-powered systems.

[Dan Martuszewski @danmartuszewski]

Worth going through the prompts it uses

[Dan Martuszewski @danmartuszewski]

Claude workflows are pretty robust

[Dan Martuszewski @danmartuszewski]

@ClaudeDevs Unfortunately, it ate a huge amount of Usage credits too!

[Dan Martuszewski @danmartuszewski]

@ClaudeDevs Thanks! Was surprised by the token usage since the reset on Friday.

[Dan Martuszewski @danmartuszewski]

May 26 AI Usage Value

$5.6k
7.39B Tokens

Claude ████████████ $2.7k
Codex █████████████ $2.9k

Claude 48.3%
Codex 51.7%

Based on @cc_usage

[Dan Martuszewski @danmartuszewski]

Codex is persistent!

[Dan Martuszewski @danmartuszewski]

Interference is expensive.
Visual changes comparison burns tokens like crazy, while it still not spot all the important differences.

[Dan Martuszewski @danmartuszewski]

And today it's back to the previous form. They listed xD

[Dan Martuszewski @danmartuszewski]

UX downgrade by @claudeai
Before: content on page - no scrollbar
Now: Popup with scrollbar even having plenty of space, I have to scroll to see everything

[Dan Martuszewski @danmartuszewski]

https://t.co/yoVFHNki7n

[Dan Martuszewski @danmartuszewski]

Spent some time during the weekend wrangling Agents HQ into a better state.

@claudeai Code agents now ping the dashboard when they need you. A red dot lands on the agent in every view, the favicon flips, the tab title counts, and a bell entry drops into the activity log so you never hear a sound and wonder what fired it.

Other bits:
- NOTIFY is now checkboxes, pick any combo of attention / crashes / status
- separate setting for whether 60s idle waits count as "attention" or not
- layout switcher collapsed to a dropdown so the header stops wrapping
- theme button is just a glowing color dot now

Small stuff, but the kind of thing that makes the difference between glancing at the dashboard and actually trusting it.

https://t.co/VlMcWfjePQ

[Dan Martuszewski @danmartuszewski]

@miroburn Jak na 1M context window zaczął się dziwnie zachowywać, przełączyłem na 200k (kilka tygodni temu). Bez większych problemów działa. Sporadycznie o coś dziwnego zapyta.

[Dan Martuszewski @danmartuszewski]

Proper Friday ending @claudeai

[Dan Martuszewski @danmartuszewski]

This is madness,
I have a skill that I manually run in the Claude Code and part of it is
cd /path/to/metadata-generation-agent && \
CLAUDECODE= claude --dangerously-skip-permissions -p "Use /seo slug-of-the-page”

With that change I will be charged separately for running it this way.

[ClaudeDevs @ClaudeDevs]

Starting June 15, paid Claude plans can claim a dedicated monthly credit for programmatic usage.

The credit covers usage of:
- Claude Agent SDK
- claude -p
- Claude Code GitHub Actions
- Third-party apps built on the Agent SDK

[Dan Martuszewski @danmartuszewski]

This is madness,
I have a skill that I MANUALLY run in the Claude Code and part of it is
cd /path/to/metadata-generation-agent && \ CLAUDECODE= claude --dangerously-skip-permissions -p "Use /seo slug-of-the-page”

With that change I will be charged separately for running it this way.

[Dan Martuszewski @danmartuszewski]

After I got banned on Claude Code few months ago (one of the hypothesis was using Claude Web in early beta on different location than I normally do) I've never tried new Claude Code remote access, as I was afraid it can happen again.
After Codex added it 8hrs ago, I tried immediately without any resistance.
The bad impression they made with those undeserved automated bans @Claude does make people less likely to use all the features they introduce.

[Dan Martuszewski @danmartuszewski]

Ultimate way to let your coding agents connect to services without exposing API keys to them

npx skills add danmartuszewski/export-key

https://t.co/5V4VlCK0U4

[Dan Martuszewski @danmartuszewski]

Did you know Claude Code cannot operate Chrome on some pages?