Peter Attia was on 60 minutes recently, so I thought it was a good time again to post some of his recommendations.

I also thought these notes on his book were great: https://www.grahammann.net/book-notes/outlive-peter-attia

Some takeaways to highlight:

• Protein: 1.6 g/kg/day at min, but 2.2 g/kg/day (or 1 g/lb of body weight) is even better
• 50g fiber per day
• Omega 3 (DHA) supplements
• Vitamin D supplementation
• Dry sauna 4x per week, 20 minutes per session, at 179F (82C) or higher (Alzheimer’s prevention)
• Pay attention to grip strength, concentric & eccentric loading, pulling motions (pull-ups and rows) and hip-hinging movements (like deadlift and squat, step-ups, hip-thrusters, single-leg variations)
• Aim for dead-hang from pull-up bar for minimum 2 minutes
• Breath training
• One-leg balance with eyes closed
• No alcohol close to bedtime
• Avoid eating hours before bedtime
• Colonoscopy at age 40 (earlier if higher risk) and then repeat every 2-3 years depending on findings
• Avoid hearing loss/be fast to introduce hearing aids if so (contributes to cognitive decline)

Here is a version of Peter Attia’s Longevity Screening Schedule by Age (via ChatGPT) – more research is needed!

Peter Attia – Longevity Screening Schedule (By Age)

Medicine 3.0 Early Detection Framework

Age 20s — Baseline & Prevention

Metabolic Health

• Fasting glucose
• Fasting insulin (calculate HOMA-IR)
• HbA1c
• Lipid panel (LDL, HDL, triglycerides)
• ApoB (baseline)
• Oral Glucose Tolerance Test (OGTT) if signs of insulin resistance
• Try Continuous Glucose Monitor (CGM) for 2–4 weeks

Fitness & Body Composition

• VO₂ max test (baseline)
• DEXA scan (lean mass, fat distribution) every 3–5 years
• Grip strength test

Cancer & General Screening

• Annual full-body skin exam
• Pap + HPV test (women)

Additional Health Data

• Sleep apnea screening if symptoms
• Vitamin D
• Thyroid panel
• Baseline audiogram

Age 30s — Detect Silent Early Disease

Metabolic / Cardiovascular

• ApoB every 3–5 years
• OGTT every 3 years
• Periodic CGM use
• hs-CRP
• Blood pressure <120/80
• If family history of heart disease: first CAC scan at ~35

Cancer

• Annual full-body skin exam
• Pap/HPV testing (women)
• Early PSA baseline for men with strong family history

Fitness / Frailty

• VO₂ max every 3–5 years
• DEXA every 3 years
• Grip strength benchmarks
• Maintain structured strength training

Other

• Sleep apnea testing if daytime sleepiness
• Annual liver + kidney panel

Age 40s — Highest-ROI Decade

The Two Low-Hanging Fruit

• Colonoscopy at age 45
• Coronary Artery Calcium (CAC) scan at age 40–45

Cardiovascular Screening

• ApoB annually
• Lp(a) once in life
• CAC every 3–5 years
• Consider Coronary CT Angiography (CTA) if:
  • CAC > 0
  • High ApoB
  • Strong family history of cardiovascular disease

Metabolic Health

• OGTT every 1–2 years
• Fasting insulin + fasting glucose yearly
• A1C (supporting test)
• 2–4 weeks of CGM annually or every other year

Cancer Screening

• Colonoscopy at 45
• Mammogram annually (women)
• Breast MRI for high-risk women
• PSA annually or every 1–2 years (men)
• Annual dermatology skin exam

Fitness / Strength / Frailty

• VO₂ max every 2–3 years
• DEXA every 2 years
• Annual grip strength
• Strength benchmarks for fall-risk prevention

Sleep

• Sleep apnea test at least once

Age 50s — Aggressive Prevention Phase

Cardiovascular

• CAC scan every 3 years
• ApoB annually
• CTA if unclear risk
• Baseline echocardiogram

Metabolic

• OGTT yearly
• Fasting insulin, glucose yearly
• A1C annually
• CGM if metabolic markers worsen

Cancer

• Colonoscopy (every 10 years or every 3–5 if polyps)
• PSA annually (men)
• Mammogram annually (women)
• Breast MRI for high-risk
• Annual dermatology exam
• Low-dose CT lung scan for ≥20 pack-year smoking history

Hormones & Organ Function

• Thyroid panel annually
• Liver and kidney function annually
• Testosterone/estradiol if symptomatic

Functional Health

• DEXA every 1–2 years
• VO₂ max every 1–2 years
• Sarcopenia monitoring: lean mass, strength, gait speed

Age 60s+ — Maintain Function & Detect Disease Early

Cardiovascular

• CAC scan as needed
• CTA more routinely if ApoB elevated
• Echocardiogram every 3–5 years
• Aortic aneurysm ultrasound (especially for men with smoking history)

Metabolic

• OGTT annually
• Fasting insulin/glucose annually
• CGM 1–2 times per year
• A1C annually

Cancer

• Colonoscopy based on prior results
• Annual PSA (men)
• Annual mammogram (women)
• Annual dermatology exam
• Annual or biennial CT lung scan if past smoker

Bone, Frailty & Function

• DEXA annually
• Grip strength annually
• Gait speed testing
• Balance testing
• Sarcopenia and fall-risk assessment

Cognitive & Sensory Health

• Cognitive testing every 1–2 years
• Annual audiogram (hearing strongly tied to cognitive decline)

Additional clips from 60 minutes:

Update Nov 20: Study on risks of Ultra Processed Foods (UPF)

https://www.bbc.com/news/articles/cy4pjjzd784o

Action is needed now to reduce ultra-processed food (UPF) in diets worldwide because of their threat to health, say international experts in a global review of research.

They say the way we eat is changing – with a move away from fresh, whole foods to cheap, highly-processed meals – which is increasing our risk of a range of chronic diseases, including obesity and depression.

Writing in The Lancet, the researchers say governments need “to step up” and introduce warnings and higher taxes on UPF products, to help fund access to more nutritious foods.

Additional therapies I’m monitoring:

• FDA-approved peptides like GLP-1 agonists
• Personalized medicine – AI gut analysis, other recommendations
• Superhuman Protocol: This protocol integrates pulsed electromagnetic fields (PEMF), exercise with oxygen training (EWOT), and photo-biomodulation (PBM) to enhance well-being. link2
• Therapeutic-plasma-exchange , link2
• Peptides, link2, link3
• Myostatin inhibitors, link2
• CBL-514

0:00 Introduction & Going Live 0:43 Meet Palmer: From Baseball to Blockchain to AI 2:24 Overview: AI Development Workflows 3:57 The New Project Setup Document 5:02 Document #1: Project Overview 5:22 Document #2: User Flow 5:45 Document #3: Tech Stack (Most Important) 6:03 Document #4: UI and Theme Rules 6:22 Document #5: Project Rules 7:17 Document #6: Phase Documentation 8:05 Docs Folder Structure 12:36 Cursor Rules vs Cursor.rules File 14:27 User Rules Configuration 18:02 Q&A: Boilerplate vs Starting Fresh 18:50 Custom Modes in Cursor 23:35 Deep Dive: New Project Setup Steps 27:17 AI-First Codebase Design 29:37 Iterative Development with Checklists 35:08 Planning Outside the IDE 38:00 Building with Documents 39:50 Context Management 43:01 Rule Inclusion Hierarchy 45:15 Cursor Notepads Feature 49:30 Chat Management 52:02 Debugging Strategies 54:46 Learning from Errors 57:23 Checkpoint Management 58:51 Closing & Resources

Slides: https://docs.google.com/presentation/d/1kCNuSck8sRpeyaPg1ElgRsMXvweU9XfL1SjO1xUi9DQ/edit?usp=sharing

https://x.com/wadefoster/status/1930680089651425452  “Zapier is measuring PMs on AI fluencyYou only get Capable if you have good prompts & use AI for PRDs and research synth. This is the baseline

AI notes:

Based on the detailed text content about Shopify’s AI-first approach, here is a comprehensive list of practices you can apply at your company to leverage AI effectively and transform your engineering and broader teams:

1. Early and Broad AI Tool Adoption:
   Shopify was an early adopter of AI coding tools like GitHub Copilot (even before commercial availability), and later introduced tools like Cursor and Cloud Code. They experiment with multiple AI tools, evaluating what works before scaling usage.
2. Cross-Functional AI Usage Beyond Engineering:
   AI tools are not limited to engineers; teams in finance, sales, support, and other functions use AI (e.g., Cursor) to build personal productivity tools like MCP servers connecting to Salesforce, Gmail, Slack, empowering non-technical staff to “vibe code” solutions independently.
3. Pairing and Collaborating with AI Labs:
   Shopify’s leadership, including the head of engineering, actively pairs with engineers from AI labs (Anthropic, OpenAI) to understand emerging tools and incorporate best practices, fostering two-way learning and innovation.
4. Building Internal AI Infrastructure (LLM Proxy & MCP Servers):
   They built an internal Large Language Model (LLM) proxy to securely access AI APIs without leaking sensitive company or customer data and manage token usage with visibility and cost controls. They also deploy many MCP (Modular Control Protocol) servers to expose internal and external data sources for AI consumption, enabling easy integration.
5. No Cost Limits on AI Usage:
   Shopify imposes no cost constraints on AI token consumption for engineers or teams, viewing AI tools as productivity investments rather than expenses to be cut. They celebrate high AI usage as a sign of impactful work.
6. Promoting AI Reflexive Culture Through Hiring:
   Shopify massively expanded its intern program (up to 1,000 interns/year) to onboard AI-native talent who naturally integrate AI tools into workflows and help shift company culture toward AI-first thinking.
7. Encouraging Hands-On Leadership and Role Modeling:
   Leaders actively use AI tools themselves and share their workflows and learnings openly, demonstrating the value and encouraging adoption. For example, engineering directors and above are required to do coding interviews, often using AI tools, to emphasize staying close to technology.
8. Allowing Non-Engineers to Build with AI (“Vibe Coding”):
   Non-technical employees are empowered to build small tools using AI without formal engineering intervention, fostering innovation and reducing bottlenecks.
9. Using AI to Reduce Toil and Address Tech Debt:
   Engineers leverage AI to tackle refactoring, tech debt, and infrastructure improvements—tasks traditionally difficult to prioritize due to feature pressures—amplifying engineers’ productivity and system quality.
10. Establishing AI-Enhanced Project Management Tools:
    Shopify built its own project management system (GSD – Get Stuff Done) that integrates AI to automatically draft weekly updates using data from pull requests and Slack conversations, reducing manual reporting toil while maintaining accountability.
11. Setting Clear AI Usage Expectations:
    A company-wide memo established that while AI use is not mandatory, the expectation is that employees should use AI tools to achieve productivity benchmarks, similar to how tools like Excel are assumed available.
12. Transparency and Measurement of AI Impact:
    Usage metrics like token consumption, exception counts, and engineering focus time are tracked and surfaced to managers to monitor AI adoption and team health.
13. Adopting a “Try and Learn” Mindset with AI Tools:
    Shopify encourages experimentation—trying new AI tools or workflows, even if initial results are imperfect or wasteful, with a willingness to discard failed attempts but learn fast.
14. Investing in AI Tooling and Infrastructure Continuously:
    The AI platform and tooling are evolving weekly, with dedicated teams managing AI infrastructure like the LLM proxy and MCP servers, ensuring security, scalability, and ease of use.
15. Combining AI with Human Oversight:
    Despite AI-generated code or content, human review and understanding remain essential—engineers must understand and verify AI outputs before submitting or deploying.
16. Encouraging AI Literacy at All Levels:
    From interns to senior engineers and managers, everyone is encouraged to develop AI skills, with internal prompt libraries and hackathons to share knowledge and accelerate adoption.
17. Hiring and Promoting Leaders Who Embrace Coding and AI:
    Engineering leaders are expected to stay technically engaged and fluent in coding, using AI tools effectively, rather than moving away from hands-on work.
18. Using Specialized AI Tools for Different Tasks:
    Shopify recognizes different AI tools have different strengths (e.g., Gum Loop for web scraping, Cloud Code for coding) and encourages using the right tool for the right job rather than one-size-fits-all.

Based on the provided text, here is a detailed list of all the AI tools, frameworks, platforms, and related technologies mentioned that Shopify uses or has experimented with:

1. GitHub Copilot – Early AI coding assistant integrated into VS Code, used extensively by Shopify engineers since 2021.
2. Cursor – An AI tool/platform used internally by Shopify, growing beyond engineering into teams like finance, sales, and support for building personal productivity solutions. Used alongside VS Code.
3. Claude Code – An agentic AI workflow tool that Shopify adopted and explored deeply, including pairing with engineers from Anthropic to understand internal uses.
4. Anthropic’s Models and Tools – Includes Anthropic’s AI models and Claude code usage; Shopify actively pairs with Anthropic engineers.
5. OpenAI Codex – Referenced as an AI code generation and refactoring tool used for tech debt reduction and infrastructure improvements.
6. Gum Loop – A browser-based automation platform used for tasks like web scraping and monitoring website content, distinct from code-centric tools.
7. LLM Proxy – Shopify’s internally built secure proxy to access various language model APIs (e.g., ChatGPT, Gemini, Claude) while protecting sensitive data.
8. Libra Chat – An open-source native chat product that Shopify contributes to and builds upon for internal AI-powered chat interfaces.
9. MCP Servers (Modular Control Protocol) – A protocol and framework Shopify uses to expose internal and external data/services for AI integration; Shopify runs about two dozen MCP servers (e.g., for Salesforce, Figma).
10. SonarQube (Sonar) – Industry-standard code quality and security analysis tool, including advanced security features, used at Shopify.
11. Work OS / OKIT – Work OS is mentioned as a SaaS app platform that includes OKIT, a user management solution, which Shopify and other startups use.
12. Static (Statsig) – Unified platform for feature flags, analytics, experiments, and more; mentioned as a tool to ship and measure product features.
13. VS Code – The primary integrated development environment (IDE) at Shopify, integrated with AI tools like GitHub Copilot.
14. Devon – An AI tool Shopify experimented with for coding workflows, though results were mixed.
15. Gemini – An AI model/provider mentioned alongside Anthropic, OpenAI, and Claude as part of the AI ecosystem Shopify engages with.
16. Claude – AI model/platform from Anthropic, used and integrated via Shopify’s LLM proxy.
17. Perplexity – Referenced as an internal “internal perplexity” style interface via Libra chat, providing advanced search and knowledge access.
18. Prompt Libraries – Shopify maintains internal prompt libraries for sharing and reusing AI prompts across teams.
19. GSD (Get Stuff Done) – Shopify’s internally built project and product management tool that integrates AI to auto-generate weekly updates using PRs and Slack conversations.
20. Slack – Used as a communication and collaboration channel; AI tools integrate with Slack for context and updates.
21. Jira and Linear – Mentioned as common project management tools Shopify consciously avoids in favor of their own GSD tool.
22. ChatGPT – Widely referenced AI chatbot platform from OpenAI, used in various capacities and integrated securely through LLM proxy.
23. Operator – Mentioned in the context of recording browser sessions to analyze commerce flows, used in AI experiments internally.

Best Practices

1. Experimentation and Iteration
   Companies and engineers should actively experiment with AI tools, sharing tips and tricks internally to discover what works best, as seen in startups like incident.io where teams discuss and refine AI usage collaboratively.
2. Integrating AI with Existing Workflows
   Successful adoption involves embedding AI tools directly into developers’ workflows, such as integrating AI into IDEs, terminals, or code review systems, ensuring seamless use rather than forcing separate tools.
3. Use Well-Defined Tasks for AI Agents
   AI agents perform better with well-scoped, well-defined tickets or tasks, allowing them to generate first passes that developers can then refine, improving efficiency and reducing errors.
4. Automate Repetitive and Low-Value Tasks
   Automation of ticketing, code generation, PR writing, and documentation helps save time and reduces cognitive load, increasing productivity as reported by engineers at Amazon and others.
5. Adopt API-First and Modular Architectures
   Following Amazon’s example, exposing all functionality via APIs enables easier integration with AI agents and automation tools, facilitating smoother AI adoption and extensibility.
6. Maintain Cautious and Incremental AI Deployment
   Big tech like Google emphasizes a slow, careful rollout to build trust and reduce mistrust among engineers, ensuring AI tools are reliable and useful before wide-scale adoption.
7. Encourage Cross-Team Sharing and Learning
   Internal communities sharing AI experiences and strategies foster faster learning and better use of AI tools, promoting cultural acceptance and knowledge dissemination.
8. Balance AI Use with Human Review
   Due to hallucinations and bugs in AI-generated code, continuous human oversight remains critical; engineers still review and correct AI output rather than fully relying on it.
9. Focus on Practical ROI and Time Savings
   Measuring realistic productivity gains (e.g., 3-5 hours a week saved) helps assess the true impact of AI tools rather than inflated expectations, guiding better investment and usage.
10. Treat AI as a New Abstraction Layer
    Recognize that LLMs and AI agents represent a lateral shift across the software stack, enabling new ways to interact with code, data, and tools at all abstraction levels.

Tools

1. Claude Code (by Anthropic)
   A command-line interface tool that sees heavy usage within AI dev tool startups, supporting terminal-based coding and allowing high portions of code to be AI-generated.
2. MCP (Model Context Protocol)
   An open protocol allowing AI agents and IDEs to connect with databases, GitHub, Google Drive, and other services, enabling chat-based querying and automation of workflows.
3. Windsurf (AI IDE Editor)
   An AI-powered editor where up to 95% of the code is generated or assisted by AI tools, supporting coding through agent interaction and passive tabbing.
4. Cursor
   An AI coding assistant with approximately 40-50% code generation success, representing a more balanced but honest approach to AI coding productivity.
5. Google’s Internal Tools
   • Cider: Google’s cloud-integrated development environment built on a VS Code fork.
   • Critique: AI-powered code review feedback tool.
   • Code Search with LLM support: Advanced search tool enhanced with AI for querying codebases.
   • Notebook LM & LM Prompt Playground: AI chat-based documentation and prompt experimentation tools.
   • MoMA Search Engine: Knowledge base powered by LLMs for internal information retrieval.
6. Amazon Q Developer Pro
   An internal AI tool specialized for AWS-related development, praised for its contextual awareness and coding assistance within the AWS ecosystem.
7. AI Agents for Automation
   Tools that automate ticketing, emails, internal workflows, and code deployment pipelines, heavily used within Amazon and other organizations for increasing efficiency.

https://www.amplifypartners.com/blog-posts/the-agent-first-developer-toolchain-how-ai-will-radically-transform-the-sdlc

• Features for an agent-first IDE
  • An intent specification layer. Agents don’t just need prompts—they need structured, machine-readable intent representations. The IDE must support DSLs for describing desired outcomes, tools to define constraints like budget limits and performance SLAs, and the ability to compare current system state to desired goals, and let agents plan accordingly.
  • Agent-framework integrations. Instead of plugins for tools like linting and debugging, future IDEs will integrate with agent frameworks—systems like Cody, Devin, Claude Code, or custom LLM pipelines. These integrations will support UIs for configuring agent roles and permissions, interfaces for visualizing and steering how different agents (e.g. frontend, backend, infra) are collaborating, and persistent agent memory that allows long-term project context across sessions.
  • Agent monitoring and telemetry. Agents need constant monitoring to ensure correctness, safety, and alignment. We will need live dashboards to visualize agent tasks and status, trace-based debugging to step through an agent’s reasoning chain and code generation path, and alerts on unexpected behaviors like unsafe code or violating performance bounds.
  • Explainability and decision logging. In a world of autonomous agents, trust depends on transparency and fine-grained auditability; more on this below. Agents will need to output action logs for each input, output, and rationale. We will want interactive code diff explanations (“why did this code change?”), plus some basic system summaries: the ability for agents to overview architecture, data flows, etc.
  • Automated testing & verification. This is the subject for an entire other blog post, but the IDE of the future needs to be able to verify agents. This verification needs to cover code tests (unit, integration, etc.), and support simulated sandbox environments to validate code behavior before merging. And finally, IDEs will need to support formal verification methods like TLA+ and ROCQ.
  • Live collaboration and feedback loops. For humans and agents to work well together, we’ll need real-time synchronization via real-time agent-human chat (like Copilot chat, but system wide), editable simulations that let you preview a feature and leave feedback (like in Google Docs or Figma), and live agent tuning via adjusting hyperparameters, toolsets, or constraints.
  • Knowledge graph and memory integration. Agents will need structured knowledge about the system, business logic, and users. I expect to see things like a semantic project graph, where the codebase (functions, models, services) is represented as an interconnected knowledge graph. We will also need embedded documentation plus persistent long-term memory that allows agents to remember past decisions, tradeoffs, and user preferences.
  • A pluggable execution environment. The IDE of the future needs a sandbox where agents can safely run and test code. Examples include containerized runtime environments, policy enforcement engines to restrict what agents can deploy or access, and a tooling-calling interface layer that lets agents use CLI tools, APIs, and libraries with traceable logs.
• Think of this new IDE as a mission control center for intelligent systems; the developer becomes a strategist, curator, and verifier rather than a code-writer.

Key Insights

•  Human-AI Collaboration Redefines Work: The future of work will revolve around humans deploying, managing, and reviewing AI agents rather than executing every task manually. This represents a fundamental shift in how knowledge work is performed, requiring new skills focused on orchestration and quality assurance rather than rote execution. The probabilistic nature of AI means human oversight remains crucial to avoid critical errors, emphasizing a hybrid human-AI workflow.
•  AI as Capability Expansion, Not Just Cost-Cutting: Box’s approach contrasts with companies focused purely on headcount reduction. Instead, AI is used to amplify what employees can accomplish, enabling faster delivery of products, more tailored marketing campaigns, and enhanced customer support. This mindset creates a virtuous cycle where efficiency gains lead to reinvestment and growth, rather than mere downsizing.
•  Jevons Paradox Applied to Labor: The speaker applies the Jevons paradox—where efficiency gains can increase overall consumption—to labor. By making tasks more efficient, AI can increase demand for services that were previously too costly or slow, such as healthcare consultations or legal reviews, leading to growth in employment in those sectors rather than contraction.
•  Enterprise AI Adoption is a Marathon, Not a Sprint: Despite rapid AI innovation, real-world organizational adoption is slow and complex due to change management challenges. Companies may take years to fully integrate AI agents into workflows, similar to how cloud adoption took decades to mature. This highlights the importance of patience and strategic planning in AI deployments.
•  New Markets and Industries Will Be Digitized: Many professional services industries have historically been under-digitized due to the nature of their work. AI agents will catalyze the digitization of sectors like legal services, healthcare, education, and drug discovery, creating new markets that were previously inaccessible to software solutions. This opens vast opportunities for startups and incumbents alike.
•  Ecosystem Diversity and Interoperability are Key: The future AI landscape will not be dominated by a single player or approach. Instead, a diverse ecosystem of AI model providers, SaaS vendors, and agent frameworks will coexist. Interoperability standards such as ADA and MCP will be critical to enable agents to communicate and collaborate across platforms, enhancing flexibility and innovation.
•  Ethical and Governance Challenges Are Front and Center: Responsibility for AI agent outputs remains with humans for the foreseeable future. As AI systems become more autonomous, new frameworks for liability and governance will emerge, possibly holding AI providers accountable. Additionally, concerns about AI-generated misinformation and the amplification of echo chambers require proactive strategies to maintain trust and information quality.
•  The Future of Software UX and Agent Integration: While some speculate that AI agents might replace traditional user interfaces, the speaker believes that most users still prefer deterministic, dashboard-style interfaces for routine interactions. Agents will operate in the background or within SaaS platforms, handling complex tasks and workflows, while users engage with familiar interfaces enhanced by AI.
•  AI Memory and Personalization Pose Complex Trade-offs: Integrating personal AI memories with corporate AI systems raises significant privacy, governance, and IP challenges. Maintaining separation (“church and state”) between personal and work agents may limit the full potential of personalized AI but is necessary to manage risks. The industry may benefit from open standards around agent memory portability to facilitate interoperability.
•  Human Creativity and Judgment Remain Irreplaceable: Despite AI’s ability to generate text, code, and other outputs, humans still need to critically assess, edit, and contextualize AI-generated content. The speaker continues to write and think independently to maintain understanding in a rapidly evolving landscape, underscoring that AI is a tool to augment, not replace, human insight.
•  Startups Have a Unique Window of Opportunity: Unlike horizontal AI assistants dominated by large players, many newly digitized industries have no entrenched incumbents, giving startups an opportunity to capture market share with innovative AI-driven solutions. This dynamic mirrors past waves of disruption seen with companies like Uber and Airbnb and suggests a fertile environment for AI entrepreneurship.
•  AI Agents Will Drive Massive API Growth and Browser Use: The rise of AI agents will fuel a surge in API creation as software vendors expose more programmable interfaces to enable agent integration. However, some tasks will always require browser-like interactions where APIs cannot fully capture the needed complexity, ensuring agent-based browser automation remains relevant.
•  Managing AI-Generated Content Quality is a Critical Challenge: The proliferation of AI-generated content risks overwhelming information ecosystems with low-quality or misleading material (“AI slop”). Verifying human-generated content and mitigating echo chamber effects will become essential, especially as agents mediate more of our interactions with digital information.
•  Change Management and Adoption Pace Are the True Bottlenecks: The speaker emphasizes that while AI technology evolves at a breakneck speed, the real constraint is how quickly individuals and organizations can adapt workflows and mental models to incorporate AI effectively. This process involves iterative experimentation, learning, and cultural shifts that take significant time.

I compared Mark Carney‘s speech ‘Growth in the age of deleveraging‘ (2011) with the Liberal party platform, using AI (o3), with some edits and added charts below:

• Bottom Line: Mark Carney’s 2011 speech called for deleveraging before the next shock hit. Instead, every sector has levered further in the intervening 14 years. That makes the Liberal platform’s reliance on larger deficits a higher-stakes wager than it would have been in 2011: the payoff must be faster productivity growth, or Canada enters the next downturn with far thinner buffers than the Bank of Canada once recommended.
• The platform emphasizes sovereignty, housing affordability and the energy transition — objectives that are hard to achieve quickly without public balance-sheet support.
• Reframing the risk. Carney’s 2011 concern was too much household leverage financed at teaser rates. The platform recasts the main risk as external shocks (U.S. tariffs, geopolitical fragmentation) and argues that public investment can crowd in private capital and lift the growth denominator faster than the debt numerator.

Snapshot of the two documents

Where the two visions align

1. Productivity & investment are Canada’s long-run growth lever.
   – The 2011 speech urges business to launch a “virtuous circle of increased investment and increased productivity.” Bank of Canada
   – The 2025 platform likewise chases capital formation—just via heavy federal spending and targeted sector plans. Liberal Party of Canada
2. Trade diversification beyond the United States.
   – Carney (2011) tells firms to “access fast-growing emerging markets.” Bank of Canada
   – The platform proposes east-west trade corridors, Arctic ports and high-speed rail to cut internal costs and open global routes. Liberal Party of Canada
3. Acknowledgement that demographics and productivity trends have turned less favourable and need policy attention. Bank of Canada Liberal Party of Canada

Where they diverge

A concise synthesis

Back in 2011, Governor Carney argued that the post-crisis world demanded prudence and private-sector-led rebalancing: pay down debt, lift business investment, find new export markets, keep the state’s balance-sheet dry.

In 2025, the Liberal platform he now fronts adopts a nearly mirror-image response to a different shock: large-scale public investment and a more interventionist federal hand to defend sovereignty, accelerate clean growth, and cushion households.

Both documents share a core belief that productivity and diversified trade are Canada’s path to prosperity. What has flipped is the chosen vehicle—fiscal activism vs. fiscal restraint—and the perceived threat—external tariffs today vs. global leverage yesterday.

Debt Metric Recap:

The alpha, your job as an AI practitioner, is to ask “What will everyone be doing in 1yr that I can do now?”

•Force yourself to use AI, think of how you can automate something when you’re doing it

•Don’t go for your old routine. Use AI and Low/NoCode. It won’t be convenient at first, but it’ll force you to spend most of your time doing what actually matters in the future?

•What’s your AI aha moment? ChatLLM? NotebookLM? Illuminate? Gemini 1.5 + Deep Research? Cursor? Bolt? ChatGPT? Etc…•

“If you’re an …. your goal should be to move to the forefront of what is made possible by AI and just ride that wave. A time like now when something is young and is blowing up but experts are few and best practices are ill-defined are when you can find absurd amounts of alpha. Honestly, the delta between what the current ….. knows and the forefront of AI is smaller than most assume, but I’m shocked at how few …… are working that direction. …. are too burned out from hype cycles that didn’t come to pass, but this time is different.”

“There is a MASSIVE CHASM between ai native and non ai native people that will be filled with LEARNING more than software.”

“”Vibe knowledge work” could mean a way of working where you rely on intuition, creativity, and AI assistance to manage, process, or generate knowledge, rather than getting bogged down in rigid structures or manual effort. Imagine using natural language to direct AI tools to research, summarize, or connect ideas for you—focusing on the big picture or the “vibe” of what you’re trying to achieve, like understanding a topic deeply or crafting a compelling argument. The AI would handle the tedious details (like sifting through data or formatting reports), while you steer it with high-level intent, maybe even through casual prompts or voice commands.”

@rowancheung

•Sam Altman on what you need to do to survive in the age of artificial intelligence.

•”You are about to enter the greatest golden age of human possibility…”

•To thrive in that world, the skills that matter most are:

•– Deep familiarity with the tools

•– Staying abreast of changes

•– Developing a great intuition for AI tools, where things are going, and how to make use of it

•– Resilience and the ability to learn things fast and evolve yourself with technology

•I know most of this stuff is a pretty big no-brainer for anyone paying attention, but here’s the takeaway:

•AI upskilling and keeping up with AI are possibly the most important skills in the world right now.

•And the most fascinating part is that new AI developments and tools come out so fast that everyone is constantly learning.

@Francis_YAO_

•I’ve been asked by few first year PhD about how to start LLM research on X, say long context modeling. My number one suggestion — though it seems a bit of unconventional — is *not* to read any papers related to long-context, but to talk to the model – Talk to the model about a text book, course slides, financial reports, novels, nonfictions, any long document you could find – Talk to the model for two whole weeks, from the morning first thing after opening up the laptop, to the evening last thing before going to the bed. – Ask every single question you could imagine, what is PCA? How does it compare to SVD? Which part of the book describes the two? What the book says exactly? – Talk to all the models you could access, GPT, Gemini, Claude, Llama … – Keep talking to the model for two whole weeks, no research, no paper, no arxiv, just talk to the model. – During the above process, continuously observe how the model behave, discover their problems, and think about why models could behave that way I found people who have gone through the above process have a fundamentally different level of understanding than people who just read papers

•If you want to do research on AI, or figure out how it can be used in your organization, the first step is talk to the models a lot. Use it for everything you do (within legal & ethical bounds). You don’t know what it does until you use it.

@johnrushx

•The Future of <Software Developer> profession:

•> software developer jobs will mainly become obsolete

•> product builders will replace them & be huge

•> code itself eventually will be abstracted far away by AI builders + component/block libraries + nocode/lowcode

•> the number of solo product builders will grow from less than a million to 10s of millions.

•> 1% of the best devs will be building these platforms & legacy/corporate/critical software

•> 99% will be forced to become “product builders”

•> most will ignore this until it’s too late and they’re out of their jobs with no skills to get a new one on same pay grade

•>> The best way to prepare for the change is to become a part-time indie hacker. Don’t quit your job; build and ship pet projects in the evenings and weekends.

•When you do so, don’t go for your old routine.

•Use AI and NoCode. It won’t be convenient at first, but it’ll force you to spend most of your time doing what actually matters in the future:

•> UX

•> ideations and validation

•> content creation

•> learning to win attention

•> translating pain points into products

•> training your eye on patterns

•> mastering productivity

•> social media

https://www.oneusefulthing.org/p/innovation-through-prompting

https://www.oneusefulthing.org/p/detecting-the-secret-cyborgs?r=i5f7

Four Questions to Ask About Your Organization.

So how can leaders start to think about the rapidly advancing nature of AI? The first thing they should do is use it. No amount of reading and research can substitute for spending 10 hours or so with a frontier model, learning what it can do. After getting familiar, companies should think about the following four questions:

1.What useful thing you do is no longer valuable? AI doesn’t do everything well, but it does some things very well. For many organizations, AI is fully capable of automating a task that used to be an important part of your organizational identity or strategy. AI comes up with more creative ideas than most people, so your company’s special brainstorming techniques may no longer be a big benefit. AI can provide great user journeys and personas, so your old product management approach is no longer a differentiator. Getting a sense of what AI can do now, and where it is heading, will allow you to have a realistic view of what might soon be delegated to an LLM.

2.What impossible thing can you do now? The flip side of the first question is that you now can do things that were impossible before. What does having an infinite number of interns for every employee get you? How does giving everyone a data analyst, marketer, and advisor change what is possible? You can look at some of the GPTs my students created as inspiration.

3.What can you move to a wider market or democratize? Prior to AI, companies were often advised to put their effort into servicing their most profitable customers, but AI has greatly changed the equation. Services and approaches that were once expensive to customize have become cheap. Prior to AI, strategy consulting firms would only work for giant clients for large fees, but now they may be able to offer effective advising to a much wider range of businesses at lower costs. Custom tutoring and mentoring, once available only to the rich, may be widely democratized.

4.What can you move upmarket or personalize? At the same time, your organization’s capabilities have increased. If you were once a small marketing firm, you can use AI to punch above your weight and offer services to elite clients that were once only available from much larger firms. With giant context windows and fast answers, every customer may be able to have a personal AI agent who knows their preferences and previous interactions with the company and communicates with them according to their preferences. Figure out the most exciting thing you can do, and see if you can make it happen.

Misguided companies will see any increase in performance from AI as an excuse to lay off staff, keeping their output the same. More forward-thinking firms will take advantage of these new capabilities to both improve the lives of their employees and expand their own capabilities. This is an area where leaders have agency over the future of AI and work. A lot depends on getting it right, and fast, because it is possible we are just getting started.

https://X.com/clairevo/status/1814747787856388435

https://www.mindstone.com/programs/ai-competency

https://www.aitra.com/contact-us

@levie
AI is going to cause us to move to higher levels of abstraction of how we work. Each level of abstraction provides more leverage than the prior level, so each bit of input leads to vastly higher output.

This has happened all throughout history when there’s major technological progress, from the Industrial Revolution with mechanical automation and in the Information Age with digital automation. The work that we do today looks far different from 100 or 50 years ago respectively.

The same will be true again with AI. What we perceive is “work” today will continue to be redefined. When you can merely think of an idea to prototype and AI can generate the code, the timelines on building software suddenly alter. When you can instantly research a topic and understand it deeply without the hundred hours going down the wrong threads, you’ll move to the next task much quicker.

This will naturally change what we spend our time doing each day in almost every field. Building software will be as much about reviewing code and considering the right prompts as it is writing the code. Delivering a healthcare outcome will mean having access to every bit of research at your fingertips instantly, augmenting anything you already know. Every domain will experience a similar impact.

Skills will matter just as much as ever, but they will look different, just as skills have changed during every other technological revolution. And more people can get started in a field they’re interested in, while the experts in the field can get even more done than they could’ve before.

In just a few years, we will look back on how we used to work and be utterly surprised how long everything took to do. It will seem implausible that you had to literally do everything yourself on a computer, the thing that was invented to help automate work.

Key Themes in Papic’s Work

1. Constraints Framework: Papic’s central idea is that geopolitical forecasting should focus on the constraints that policymakers face rather than their preferences or ideologies. He argues that these constraints—political, economic, and geopolitical—are more reliable indicators of future actions than the stated goals of leaders.
   • Example: During the Greek financial crisis, despite the Greek government’s preferences for brinkmanship with the EU, the constraints imposed by the Greek median voter’s desire to remain in the eurozone ultimately dictated the outcome. This shows how domestic political constraints can override government preferences.
2. Material Constraints Over Preferences: Papic emphasizes that material constraints, such as economic realities and geopolitical pressures, are more significant than the personal preferences of leaders. This approach suggests that understanding these constraints can lead to better predictions of geopolitical events.
   • Example: In the context of U.S.-China relations, despite ideological differences and security concerns, the economic interdependence between the two countries acts as a constraint that prevents a complete decoupling. This highlights how economic constraints can limit the extent of geopolitical rivalry.
3. Multipolar World Dynamics: Papic discusses the shift from a unipolar to a multipolar world, where multiple power centers exist. This complexity requires a nuanced understanding of how different global players interact and the constraints they face.
   • Example: The ongoing tensions between the U.S. and China are influenced by the multipolar nature of global politics, where neither country can fully dominate the other due to the presence of other significant powers and economic interdependencies.
4. Political Capital and Median Voter Theorem: Papic introduces the concept of political capital, which includes factors like popularity and legislative power, and the Median Voter Theorem, which suggests that political outcomes are often aligned with the preferences of the median voter.
   • Example: The Brexit vote can be seen as a reflection of the median voter’s preference for sovereignty and immigration control over economic integration with the EU, demonstrating how political constraints can shape major geopolitical decisions.

Recent Commentary on the Canada–U.S. Trade Dispute

January 29:

• Rumors Surface: Early signs of a potential trade conflict emerge when unnamed Canadian officials express concern about new tariffs on specific Canadian exports. Markets react mildly, anticipating negotiation rather than escalation.

January 31 – from @Geo_papic

• @Geo_papic: “My highly sophisticated model on where we are in the Canada–U.S. trade dispute. This is proprietary, but I have decided to share it with my X followers. You are welcome.”

February 1:

• U.S. announces tariffs on Canada
• Canada escalates and announces counter-tariffs on the U.S.

February 2:

Source: @SpecialSitsNews

@PeterBerezinBCA “Goldman this morning: “While the outlook is unclear, we think the Canada- and Mexico-focused tariffs are likely to be short-lived.”The problem with this view is that Trump won’t change course unless the stock market sells off bigly, but the stock market won’t sell off bigly if investors continue to think that the tariffs will be lifted soon.”

February 3:

@DeItaone “SENIOR CANADA GOV’T OFFICIAL TELLS NEW YORK TIMES THAT OTTAWA IS NOT OPTIMISTIC A REAL OFF-RAMP FROM TARIFFS EXISTS FOR CANADA THE WAY IT MATERIALIZED FOR MEXICO”

@LDrogen Trudeau’s task today is to come up with something completely performative to hand Trump because there doesn’t exist anything within the reasonable universe that isn’t performative he can actually hand him

@Geo_papic “All right, I have re-run the numbers on my sophisticated constraint-based trade war model. Here are the results (color coordinated!).”

• @Geo_papic “I LOVE this statement. The White House is practically PLEADING with Ottawa to give in with a meaningless political display of kowtowing ala the 10,000 Mexican troops. I hope RoW is watching and learning. It’s not that difficult people!”
  https://www.reuters.com/world/americas/white-house-mexico-is-serious-canada-appears-have-misunderstood-trumps-executive-2025-02-03/

• @Geo_papic “When you spend all day fielding media and client calls on tariffs that won’t even move the markets over the course of 24 hours because constraints > preferences…”
• @Geo_papic “very nice call on going long the Canadian dollar today at 10:30 AM. On the Bca webcast.”

Canadian Response

February 3 – from @JustinTrudeau

@JustinTrudeau “I just had a good call with President Trump. Canada is implementing our $1.3 billion border plan—reinforcing the border with new choppers, technology and personnel, enhanced coordination with our American partners, and increased resources to stop the flow of fentanyl. Nearly 10,000 frontline personnel are and will be working on protecting the border.
In addition, Canada is making new commitments to appoint a Fentanyl Czar, list cartels as terrorists, ensure 24/7 eyes on the border, launch a Canada–U.S. Joint Strike Force to combat organized crime, fentanyl, and money laundering. I have also signed a new intelligence directive on organized crime and fentanyl, backed by $200 million.
Proposed tariffs will be paused for at least 30 days while we work together.”

This development illustrates ‘constraints over preferences’ on both sides. Canada’s political capital and economic interests propel efforts to avoid a significant trade escalation (at least it did at the last minute), while the U.S. has constraints on the potential losses from a trade war, making a face-saving agreement mutually advantageous. We shall see what happens when it comes to NATO/military spend, defense of the arctic, however,…

Why Papic’s Approach Matters for Investors and Analysts

By zeroing in on the material, political, and economic constraints that decision-makers face, Papic’s method helps observers anticipate sudden shifts—such as an unexpected tariff pause or border-control measure. These insights often reveal that market-moving developments are less about leaders’ ideologies and more about the hard realities they cannot ignore.

Broader Context on Border, Crime, FENTANYL,…

The concessions by Trudeau may help towards solving another problem…there is ongoing commentary from figures like Sam Cooper, Stephen Punwasi, and Marc Cohodes, who raise questions about illicit financial flows and potential ties to organized crime, issues that might not have been adequately addressed up until now. Its a pity it took stupid tariff threats for this to materialize

Odds & Ends – An interesting perspective on trade:

 Hudson Bay Capital: “The root of the economic imbalances lies in persistent dollar overvaluation that prevents the balancing of international trade, and this overvaluation is driven by inelastic demand for reserve assets.” “As global GDP grows, it becomes increasingly burdensome for the United States to finance the provision of reserve assets and the defense umbrella, as the manufacturing and tradeable sectors bear the brunt of the costs.”

“The emergence of a world where human intelligence is no longer a constraint on economic, social, and scientific endeavors would rapidly transform production and innovation. Cognitive hyper abundance would erase many traditional bottlenecks in R&D, enabling near-instantaneous breakthroughs in scientific fields ranging from biotechnology to clean energy. Markets would recalibrate as the marginal cost of knowledge-based products and services approached zero, dissolving old competitive barriers and creating wealth at an unprecedented rate. Economic sectors that once relied on specialized expertise would expand or shift toward tasks requiring creativity and empathy, while newly automated cognitive tasks would release immense human capacity to experiment with novel forms of entrepreneurship, research, and personal development.

Socially, the dissolution of intellectual barriers would trigger mass realignments in education, cultural expression, and governance. Education could become a process of creative exploration rather than rote instruction, with learners guided by systems capable of personalizing lessons and instantly correcting misunderstandings. The resulting democratization of advanced skills might neutralize inequality in knowledge access, though new challenges would arise around how societies regulate such transformative power. Old forms of prestige built upon scarcity of expertise could recede, and new forms of social distinction might develop around originality, emotional intelligence, and moral leadership.

In addition, political structures would likely adapt to manage the turbocharged pace of discovery. Governments might struggle to keep up with rapidly evolving norms and industries, forcing them to reimagine regulatory frameworks and possibly even the meaning of representative decision-making. The potential to solve existential challenges, including resource scarcity and environmental degradation, would be magnified by the proliferation of supercharged problem-solving tools. However, there could also be acute risks if the disparity in access and control of cognitive abundance were to concentrate power in the hands of a small elite or specialized organizations. These first-order consequences illustrate how solving ASI, defined here as removing the constraint of human intelligence, would catalyze profound shifts in virtually every dimension of human life.“