Brennan McDonald
@brennanmcdonald.bsky.social
I like reading and writing. Know some things about financial services and technology. I edit the Global Custody Pro newsletter and write a Substack.
https://www.brennanmcdonald.com
https://www.brennanmcdonald.com
Check out my year-in-review post: https://www.brennanmcdonald.com/p/operating-model-compression-a-2025
Operating Model Compression: A 2025 AI Year in Review
I thought frameworks would be enough. I was wrong. Here is a review of my 16 posts that mattered most in 2025
www.brennanmcdonald.com
December 29, 2025 at 4:07 AM
Check out my year-in-review post: https://www.brennanmcdonald.com/p/operating-model-compression-a-2025
"AI is garbage. AI doesn't work. AI always hallucinates."
I read these comments daily.
However, what's interesting is that the people saying this are working through the classic stages of grief.
And they don't even realise it.
Today's article: why AI resistance follows the same pattern a
I read these comments daily.
However, what's interesting is that the people saying this are working through the classic stages of grief.
And they don't even realise it.
Today's article: why AI resistance follows the same pattern a
July 23, 2025 at 11:14 PM
"AI is garbage. AI doesn't work. AI always hallucinates."
I read these comments daily.
However, what's interesting is that the people saying this are working through the classic stages of grief.
And they don't even realise it.
Today's article: why AI resistance follows the same pattern a
I read these comments daily.
However, what's interesting is that the people saying this are working through the classic stages of grief.
And they don't even realise it.
Today's article: why AI resistance follows the same pattern a
Regulatory reporting and compliance are a complex domain.
One of the open questions I have over the next few years is how regulators will adapt their approach as technology continues to improve.
At some point, it will make sense for regulatory entities to have read-only AI agents from regulat
One of the open questions I have over the next few years is how regulators will adapt their approach as technology continues to improve.
At some point, it will make sense for regulatory entities to have read-only AI agents from regulat
July 3, 2025 at 9:00 PM
Regulatory reporting and compliance are a complex domain.
One of the open questions I have over the next few years is how regulators will adapt their approach as technology continues to improve.
At some point, it will make sense for regulatory entities to have read-only AI agents from regulat
One of the open questions I have over the next few years is how regulators will adapt their approach as technology continues to improve.
At some point, it will make sense for regulatory entities to have read-only AI agents from regulat
There are tradeoffs in everything. It's fascinating to read a significant amount of AI safety literature that focuses very heavily on low-probability outcomes, often overlooking the possibility that any societal benefits could exist.
Economist Bryan Caplan has written extensively about concepts
Economist Bryan Caplan has written extensively about concepts
July 2, 2025 at 8:00 PM
There are tradeoffs in everything. It's fascinating to read a significant amount of AI safety literature that focuses very heavily on low-probability outcomes, often overlooking the possibility that any societal benefits could exist.
Economist Bryan Caplan has written extensively about concepts
Economist Bryan Caplan has written extensively about concepts
The use of AI in financial services will require significant effort.
AI exposes bad data, poor processes, and insufficient documentation.
If you have weak leaders and messy thinking about technology in business, you're at risk.
Most firms will need to spend a significant amount of money p
AI exposes bad data, poor processes, and insufficient documentation.
If you have weak leaders and messy thinking about technology in business, you're at risk.
Most firms will need to spend a significant amount of money p
July 1, 2025 at 9:00 PM
The use of AI in financial services will require significant effort.
AI exposes bad data, poor processes, and insufficient documentation.
If you have weak leaders and messy thinking about technology in business, you're at risk.
Most firms will need to spend a significant amount of money p
AI exposes bad data, poor processes, and insufficient documentation.
If you have weak leaders and messy thinking about technology in business, you're at risk.
Most firms will need to spend a significant amount of money p
I've tried both Gemini CLI and Claude Code.
Claude Code is much easier to work with and get the right outcome.
Hitting the limits is frustrating, but it's a known Anthropic issue.
With the use of Git and frequent intervention, most "but what about" issues can be avoided.
The value-for-
Claude Code is much easier to work with and get the right outcome.
Hitting the limits is frustrating, but it's a known Anthropic issue.
With the use of Git and frequent intervention, most "but what about" issues can be avoided.
The value-for-
June 30, 2025 at 8:30 PM
I've tried both Gemini CLI and Claude Code.
Claude Code is much easier to work with and get the right outcome.
Hitting the limits is frustrating, but it's a known Anthropic issue.
With the use of Git and frequent intervention, most "but what about" issues can be avoided.
The value-for-
Claude Code is much easier to work with and get the right outcome.
Hitting the limits is frustrating, but it's a known Anthropic issue.
With the use of Git and frequent intervention, most "but what about" issues can be avoided.
The value-for-
Many AI discussions overlook the fact that the lessons learned from project failures and technology disasters remain relevant. AI projects won't fail because of AI - they'll fail because there wasn't enough upfront thinking on how to manage the basic project and change risks.
June 30, 2025 at 2:04 AM
Many AI discussions overlook the fact that the lessons learned from project failures and technology disasters remain relevant. AI projects won't fail because of AI - they'll fail because there wasn't enough upfront thinking on how to manage the basic project and change risks.
Operating Model Compression is an idea I've been writing about this year. Because of AI, firms can turn the "10x developer" concept into a 90% reduction in resource footprint. AI tools will enable some competitors and AI-first startups to deliver the same outcome to your customers at 10x or even 100
June 24, 2025 at 10:19 PM
Operating Model Compression is an idea I've been writing about this year. Because of AI, firms can turn the "10x developer" concept into a 90% reduction in resource footprint. AI tools will enable some competitors and AI-first startups to deliver the same outcome to your customers at 10x or even 100
Self-optimising AI agents will use API endpoints to get tasks done in real-time.
No human coordination required.
open.substack.com/pub/brennanm...
No human coordination required.
open.substack.com/pub/brennanm...
open.substack.com
May 30, 2025 at 4:01 AM
Self-optimising AI agents will use API endpoints to get tasks done in real-time.
No human coordination required.
open.substack.com/pub/brennanm...
No human coordination required.
open.substack.com/pub/brennanm...
You have 5 years to compress your operating model by 90%, or an AI-first competitor will eat your lunch.
open.substack.com/pub/brennanm...
open.substack.com/pub/brennanm...
May 24, 2025 at 10:19 PM
You have 5 years to compress your operating model by 90%, or an AI-first competitor will eat your lunch.
open.substack.com/pub/brennanm...
open.substack.com/pub/brennanm...
"The same people I know who consistently missed or were late to adopt every major technology trend since the 1990s are now dismissing artificial intelligence's potential."
www.brennanmcdonald.com/p/on-ai-2027...
www.brennanmcdonald.com/p/on-ai-2027...
On AI 2027 and Vibe Coding
Superhuman AI research will be awesome until it...isn't?
www.brennanmcdonald.com
May 15, 2025 at 12:46 PM
"The same people I know who consistently missed or were late to adopt every major technology trend since the 1990s are now dismissing artificial intelligence's potential."
www.brennanmcdonald.com/p/on-ai-2027...
www.brennanmcdonald.com/p/on-ai-2027...
One thing I'm convinced of about the imminent artifical intelligence-led social upheaval:
- there won't be any retraining or upskilling schemes
- there won't be a universal basic income
- everyone is pretty much on their own here
- nothing will save you from this but capital reserves
Ouch
- there won't be any retraining or upskilling schemes
- there won't be a universal basic income
- everyone is pretty much on their own here
- nothing will save you from this but capital reserves
Ouch
May 14, 2025 at 12:01 AM
One thing I'm convinced of about the imminent artifical intelligence-led social upheaval:
- there won't be any retraining or upskilling schemes
- there won't be a universal basic income
- everyone is pretty much on their own here
- nothing will save you from this but capital reserves
Ouch
- there won't be any retraining or upskilling schemes
- there won't be a universal basic income
- everyone is pretty much on their own here
- nothing will save you from this but capital reserves
Ouch
I wrote about why more AI will deliver better, faster, cheaper outcomes.
We'll still need secure guardrails and controls - but it's here to stay and firms need to start adapting much faster to stay competitive against challenger threats.
open.substack.com/pub/brennanm...
We'll still need secure guardrails and controls - but it's here to stay and firms need to start adapting much faster to stay competitive against challenger threats.
open.substack.com/pub/brennanm...
Most AI Critics Are Wrong About Slop
Artificial intelligence will make everything better, faster, and cheaper
open.substack.com
May 9, 2025 at 6:04 AM
I wrote about why more AI will deliver better, faster, cheaper outcomes.
We'll still need secure guardrails and controls - but it's here to stay and firms need to start adapting much faster to stay competitive against challenger threats.
open.substack.com/pub/brennanm...
We'll still need secure guardrails and controls - but it's here to stay and firms need to start adapting much faster to stay competitive against challenger threats.
open.substack.com/pub/brennanm...
I’m interested in understanding the thinking behind folks still enrolling in more certifications or more higher education?
Why do people think this will be additive / net positive in this AI era?
What am I missing? Interested to hear these perspectives.
Why do people think this will be additive / net positive in this AI era?
What am I missing? Interested to hear these perspectives.
May 6, 2025 at 12:59 AM
I’m interested in understanding the thinking behind folks still enrolling in more certifications or more higher education?
Why do people think this will be additive / net positive in this AI era?
What am I missing? Interested to hear these perspectives.
Why do people think this will be additive / net positive in this AI era?
What am I missing? Interested to hear these perspectives.
I wrote a roundup of what I've covered in the global custody industry in April 2025:
www.linkedin.com/pulse/what-i...
www.linkedin.com/pulse/what-i...
What I Wrote About The Global Custody Industry In April 2025
Since leaving my corporate job at a global custodian I've been writing extensively on the global custody industry. I started an independent media publication - Global Custody Pro - and below I'm shari...
www.linkedin.com
May 4, 2025 at 2:56 AM
I wrote a roundup of what I've covered in the global custody industry in April 2025:
www.linkedin.com/pulse/what-i...
www.linkedin.com/pulse/what-i...
I wrote about how AI risk avoidance is actually risky - especially for big incumbents in regulated industries.
open.substack.com/pub/brennanm...
open.substack.com/pub/brennanm...
Stop Box-Ticking, Start Winning
Free your teams from the bureaucratic hurdles that kill innovation
open.substack.com
May 2, 2025 at 4:53 AM
I wrote about how AI risk avoidance is actually risky - especially for big incumbents in regulated industries.
open.substack.com/pub/brennanm...
open.substack.com/pub/brennanm...
I wrote about how actually reading things is an advantage when so many just want the tl;dr.
In the AI era, deep reading enables better, faster, cheaper guidance of AI tooling - you need a fact base to navigate the prompt window from.
www.brennanmcdonald.com/p/deep-readi...
In the AI era, deep reading enables better, faster, cheaper guidance of AI tooling - you need a fact base to navigate the prompt window from.
www.brennanmcdonald.com/p/deep-readi...
April 25, 2025 at 2:19 AM
I wrote about how actually reading things is an advantage when so many just want the tl;dr.
In the AI era, deep reading enables better, faster, cheaper guidance of AI tooling - you need a fact base to navigate the prompt window from.
www.brennanmcdonald.com/p/deep-readi...
In the AI era, deep reading enables better, faster, cheaper guidance of AI tooling - you need a fact base to navigate the prompt window from.
www.brennanmcdonald.com/p/deep-readi...
Interesting moves from Windsurf in the AI vibe coding space.
Got an email announcing a price decrease and a change to their “flow action” charging - a major pain point.
Unfortunately I have already switched to Roo Code + OpenRouter.
That’s the issue with the speed of change now: you have hours.
Got an email announcing a price decrease and a change to their “flow action” charging - a major pain point.
Unfortunately I have already switched to Roo Code + OpenRouter.
That’s the issue with the speed of change now: you have hours.
April 21, 2025 at 11:42 PM
Interesting moves from Windsurf in the AI vibe coding space.
Got an email announcing a price decrease and a change to their “flow action” charging - a major pain point.
Unfortunately I have already switched to Roo Code + OpenRouter.
That’s the issue with the speed of change now: you have hours.
Got an email announcing a price decrease and a change to their “flow action” charging - a major pain point.
Unfortunately I have already switched to Roo Code + OpenRouter.
That’s the issue with the speed of change now: you have hours.
Write about what you want to know in detail.
Get a frontier model to optimise your prompt for deep research in JSON format.
Then execute the prompt in a fresh chat.
This works really well for improving the output of Google Gemini Pro 2.5 Deep Research prompts.
I’m getting much better output.
Get a frontier model to optimise your prompt for deep research in JSON format.
Then execute the prompt in a fresh chat.
This works really well for improving the output of Google Gemini Pro 2.5 Deep Research prompts.
I’m getting much better output.
April 18, 2025 at 12:38 AM
Write about what you want to know in detail.
Get a frontier model to optimise your prompt for deep research in JSON format.
Then execute the prompt in a fresh chat.
This works really well for improving the output of Google Gemini Pro 2.5 Deep Research prompts.
I’m getting much better output.
Get a frontier model to optimise your prompt for deep research in JSON format.
Then execute the prompt in a fresh chat.
This works really well for improving the output of Google Gemini Pro 2.5 Deep Research prompts.
I’m getting much better output.
1:1 custom curriculum, tutoring, and learning validation is a product space that will grow a lot because of AI over the next few years.
Why pay for a mini course if you take the learning outcomes and get a deep research report for a first pass overview then use it as a base for asking questions?
Why pay for a mini course if you take the learning outcomes and get a deep research report for a first pass overview then use it as a base for asking questions?
April 16, 2025 at 11:33 PM
1:1 custom curriculum, tutoring, and learning validation is a product space that will grow a lot because of AI over the next few years.
Why pay for a mini course if you take the learning outcomes and get a deep research report for a first pass overview then use it as a base for asking questions?
Why pay for a mini course if you take the learning outcomes and get a deep research report for a first pass overview then use it as a base for asking questions?
I wrote about the AI-2027 predictions and vibe coding a risk portal for data journalism.
The train is so far away from the station it ain't returning - ever.
open.substack.com/pub/brennanm...
The train is so far away from the station it ain't returning - ever.
open.substack.com/pub/brennanm...
On AI 2027 and Vibe Coding
Superhuman AI research will be awesome until it...isn't?
open.substack.com
April 4, 2025 at 4:00 AM
I wrote about the AI-2027 predictions and vibe coding a risk portal for data journalism.
The train is so far away from the station it ain't returning - ever.
open.substack.com/pub/brennanm...
The train is so far away from the station it ain't returning - ever.
open.substack.com/pub/brennanm...
I think more people should use today’s tariff news to learn more about trade policy through reading books and journal articles instead of posting their reckons.
April 3, 2025 at 12:04 AM
I think more people should use today’s tariff news to learn more about trade policy through reading books and journal articles instead of posting their reckons.
Everything I wrote about the global custody industry in March 2025:
www.linkedin.com/pulse/everyt...
www.linkedin.com/pulse/everyt...
Everything I Wrote About Global Custody In March
Since leaving my corporate job at a leading global custodian I've been writing extensively on the global custody industry. I started an independent publication - Global Custody Pro - and below I'm sha...
www.linkedin.com
April 1, 2025 at 3:49 AM
Everything I wrote about the global custody industry in March 2025:
www.linkedin.com/pulse/everyt...
www.linkedin.com/pulse/everyt...
One thing about trade is many often forget that trade in services could just as easily attract retaliation.
March 31, 2025 at 10:42 PM
One thing about trade is many often forget that trade in services could just as easily attract retaliation.
I wrote about the "Studio Ghibli" moment and what it means.
So many people are trying out the 4o image generation model and realising every few months AI tools are getting better-and-better.
www.brennanmcdonald.com/p/on-studio-...
So many people are trying out the 4o image generation model and realising every few months AI tools are getting better-and-better.
www.brennanmcdonald.com/p/on-studio-...
On Studio Ghibli
Why OpenAI's new 4o image generation model rocks
www.brennanmcdonald.com
March 28, 2025 at 3:00 AM
I wrote about the "Studio Ghibli" moment and what it means.
So many people are trying out the 4o image generation model and realising every few months AI tools are getting better-and-better.
www.brennanmcdonald.com/p/on-studio-...
So many people are trying out the 4o image generation model and realising every few months AI tools are getting better-and-better.
www.brennanmcdonald.com/p/on-studio-...