Harvey growthproperties

Harvey growthproperties I help people live on their own terms using property. Through coach or handsfree investing.

I f**ked up, and I apologise.It’s been a challenging year, and I think it’s important to share the challenges alongside ...
26/09/2026

I f**ked up, and I apologise.

It’s been a challenging year, and I think it’s important to share the challenges alongside the wins.

One of those challenges happened recently.

70+ customers were incorrectly charged £74 for an AI agent they didn’t want.

The software wasn’t mine and the £74 wasn’t my money. It went to the software provider.

But I sold the software and collected the payment on their behalf.

So ultimately, the responsibility sits with me.

And rightly so.

I made the decision that whether the software company reimbursed me or not, I would put those customers right.

But it created a massive customer service challenge.

Alongside those 70+ cases, we have a large community where people were also making normal subscription cancellations.

Messages were coming through Instagram, Skool, WhatsApp, email, texts and phone calls.

I had a VA managing customer support, but with the sheer volume coming through so many different places, some things got missed.

So what have I done about it?

Firstly, I stopped taking on new people into the programme.

Secondly, I’m fixing the infrastructure.

I’m building an AI team member that can work across multiple platforms, bring everything together and help make sure messages don’t get missed.

For anybody affected by this, I do apologise. Please accept my apology.

It was never my intention for this to happen. However, it’s still my responsibility.

None of this is a justification.

It’s simply an explanation of what happened and what I’m doing to make sure it doesn’t happen again.

Evolve your ideas

Live on your terms

Harvey
The AI Property Guy

25/09/2026

Comment WORKSHOP and I'll DM you the link.

23/09/2026

* Give Astra a explicit usage budget: Prompt it to finish a task using a percentage cap (e.g., “complete this within 2% of my limit”).

* Set up workspace Router Files: Create lightweight .md maps (like product.md) so Astra opens exact files instead of reading the entire codebase.

* Keep reports and tool logs minimal: Tell Astra to only report detailed output on errors and give 1-line summaries when steps work.

* Turn off unused tools and plugins: Extensions, image generators, and web search integrations consume context tokens on every message, even when you aren’t using them.

* Use direct file paths when prompting: Instead of asking Astra to debug blindly, give it the exact path (Ctrl+Shift+C on Windows) to save 20%+ on search tokens.

* Enable Mid-Run Steering: Switch follow-up behavior to “Steer” in settings to stop bad runs in real time before wasting tokens on unwanted outputs.

* Maintain a progress log: For long jobs, have Astra update a .progress_log.md file so crashes don’t force it to re-do work you already paid for.

* Compress CLI outputs (RTK / Headroom): Use terminal compression layers to trim long logs and shell responses before Astra reads them.

* Skip “Max” reasoning effort: “Extra High” reasoning gives nearly identical intelligence results while saving up to 40% in token usage per task.

* Disable Fast Mode (1.5x Speed): Speed modes burn through rate limits significantly faster for the same work—turn it off unless you are in an urgent rush.

23/09/2026

* Give Astra an explicit usage budget: Prompt it to finish a task using a percentage cap (e.g., “complete this within 2% of my limit”).

* Set up workspace Router Files: Create lightweight .md maps (like product.md) so Astra opens exact files instead of reading the entire codebase.

* Keep reports and tool logs minimal: Tell Astra to only report detailed output on errors and give 1-line summaries when steps work.

* Turn off unused tools and plugins: Extensions, image generators, and web search integrations consume context tokens on every message, even when you aren’t using them.

* Use direct file paths when prompting: Instead of asking Astra to debug blindly, give it the exact path (Ctrl+Shift+C on Windows) to save 20%+ on search tokens.

* Enable Mid-Run Steering: Switch follow-up behavior to “Steer” in settings to stop bad runs in real time before wasting tokens on unwanted outputs.

* Maintain a progress log: For long jobs, have Astra update a .progress_log.md file so crashes don’t force it to re-do work you already paid for.

* Compress CLI outputs (RTK / Headroom): Use terminal compression layers to trim long logs and shell responses before Astra reads them.

* Skip “Max” reasoning effort: “Extra High” reasoning gives nearly identical intelligence results while saving up to 40% in token usage per task.

* Disable Fast Mode (1.5x Speed): Speed modes burn through rate limits significantly faster for the same work—turn it off unless you are in an urgent rush.

21/09/2026

Comment PA and I’ll send you my playbook

20/09/2026

Comment 3D and I’ll dm you the playbook

20/09/2026

Comment 3D and I’ll send you the play book

GPT 6 Astra mind blowing Building web pages and funnels in my GoHighLevel software while I’m busy working on other thing...
15/09/2026

GPT 6 Astra mind blowing

Building web pages and funnels in my GoHighLevel software while I’m busy working on other things 🤯

Address

3 Nelson Terrace
Stockton-on-Tees

Website

,

Alerts

Be the first to know and let us send you an email when Harvey growthproperties posts news and promotions. Your email address will not be used for any other purpose, and you can unsubscribe at any time.

Shortcuts

Share

Category