ActionableOps

ActionableOps is an AI consulting company that helps businesses unlock AI's potential to excel in their operations. We specialize in guiding companies through the AI landscape with services such as AI governance, Responsible AI, AI Policy development, AI Ethics, AI Training, and AI Implementation. Partnering with industry leaders like Textcortex and MindStudio, we ensure that our clients gain the knowledge and tools needed to thrive with AI.

Founder / AI Consulting

Kansas City AI Club

Primary Purpose: Foster awareness, adoption, and hands-on learning of the latest AI technologies by leading discussions, events, and workshops tailored to real-world business applications.

Key Responsibilities:
 - Technology Trends Leadership in AI
 - Curate and present emerging AI technologies (e.g., generative AI, voice agents, AI-driven automation).
 - Deliver accessible, business-focused insights on cutting-edge advancements.
 - Business Use Case Applications
 - Develop and showcase practical AI use cases (e.g., AI in customer service, operations, or marketing).

Workshops & Education
 - Organize hands-on workshops to teach members how to apply AI in real-world business settings.
 - Guide participants through implementation frameworks tailored to their industries.
 - Event Programming Support
 - Host and facilitate demos at Kansas City AI Club meetups.
 - Develop quarterly in-person or virtual workshop series, such as "AI Applications for Business Success."

Board Chair - Emerging Tech

UMKC Henry W. Bloch School of Management

The Regnier Institute Mentor in Residence Program connects seasoned mentors with student entrepreneurs in need of a helpful ear.

Mentor-In-Residence

National Speakers Association-Kansas City Chapter

Speaker

MindStudio

Agency Partner

RunWhen

RunWhen is a platform for troubleshooting complex tech stacks. It helps engineers identify when issues they are chasing lie in their app, their cloud-native platform or their vendors by answering the question: “What do I run and when?”

Advisor

Nobl9

Nobl9 is a service level objective SaaS platform that connects reliability metrics to business goals via a modern engineering workflow. Backed by CRV, Battery, Bonfire Ventures, Harmony Partners Cisco and ServiceNow.

Director of Operations

Code Ninjas

Code Ninjas is the world's largest and fastest growing in person children's franchise startup, that teaches kids how to code computers by building video games. Backed by Eagle Merchant Partners.

Director of Operations / Owner

OMT Brands

OMT Brands created a line of consumable pet products to be marketed and sold online through Amazon FBA and the Amazon, Ebay, Sears, and Walmart marketplaces.

Director of E-commerce Operations

City of Lakewood, Washington

Police Officer

🚀 Founder | Startup Operations Executive | Business Scaling Expert | Data Analytics | P&L Management | Process Optimization | Customer-Centric | Revenue Growth | Operational Efficiency | SaaS | Change Management

I’m Dale Thomas and here’s my story in a nutshell. 

- Authored the best-selling book “The Ultimate A.I. Guide for Small Business Success.” 
- Founded several technology businesses
- 20+ years of operations experience in Start-Ups and Enterprise (Ex - Microsoft) 
- Focused on helping SMB’s make sense of AI.  I specialize in AI agents, AI Consulting, responsible AI, ethical AI, policy creation, governance, and implementation of AI solutions. 

What you’ll get by following me: 
🔥Transparent & actionable AI advice
📆 Tips and Tricks Implementing AI Agents
🔍 Insights on implementing ethical AI practices
📢 Up-to-date information on AI regulations and legislation  

AI Partnerships
* TextCortex AI Partner - get a free account https://textcortex.com?via=ops 

A.I. Education
* MindStudio Certified AI Expert
* The London School of Economics - Ethics of AI 
* University of Pennsylvania - AI For Business
* Vanderbilt University - Prompt Engineering for ChatGPT
* MindStudio Certified AI Developer

Struggling to make sense of AI for your small to medium-sized business? 

I’m here to help - so hit that “Follow” button!  

Dale

Dale Thomas

David Arnoux

I help B2B founders and marketing leaders grow with AI, GTM strategy and LinkedIn.
20 years scaling companies: Growth Tribe (40,000+ learners), e-commerce brand (€30M+), SaaS (15m+ ARR).

Now:
→ Fractional CxO for AI adoption & GTM @ heyarnoux.com
→ Building AI tools for founders @ humanoidz.ai (ViralBrain, NuggetFinder, BuilderBox)
→ Running an invite-only AI community @ thegenaicircle.com

I write about AI, growth and building in public.

Co-CEO

Growth Tribe is here to empower adults to always be learning. Our solutions ensure people & organisations can acquire the digital skills they need to adapt, stay up-to-date and in the lead!

We’ve built a reputation for fast-paced, impactful & enjoyable learning, offering training & coaching in Data Science, Analytics, Artificial Intelligence, Growth & Marketing, Innovation, UX Design & Product Management. 

With campuses in 6 European countries, we have supported the learning journeys of over 16,000 adults in over 900 organisations. Our 80+ staff are dedicated to creating Europe’s best learning experiences, to empower humans to always be learning.

You've seen all the hype around Claude..CoWork...Code

You're lost. Lemme try to help a bit...

But first some business history :) There's a question that haunted management theory for decades. Why do some leaders, foudners, operators scale effortlessly while others hit invisible ceilings? Peter Drucker identified the answer in 1967. The effective executives weren't the ones who waited until they fully trusted someone before handing off work. They extended trust slightly ahead of where it had been earned. That uncomfortable gap was where organizations grew.

The same dynamic is now playing out with AI. I've spent months in a sea of watching how people actually use Claude. Whatemerged was a spectrum (or a mental model) of ten distinct levels. Level 0 is basic chat without Claude. Level 10 is autonomous agents running while you sleep (it's really happening...just search for Ralph Wiggum brute-forcing)
                              
Most people are stuck between 0 and 2. Not because the tools are limited. Because each level up requires surrendering a bit of control (and setup time). Level 3 means letting Claude hold context across weeks of work. Level 6 means walking away while it completes tasks. Level 9 means background agents making real decisions without supervision (Still super risky!). 

The people moving fastest aren't the most technical anymore. They're the ones who learned to delegate to humans and now apply the same muscle to machines. Two years ago, prompting was a competitive advantage. Today it's table stakes. The scarcity migrated. When everyone can ask questions, the advantage moves to those who can delegate answers. To a certain extent the quality of your taste/eval BUT ALSO NOT the quality of your repository now dictates the quality (and quantity) of the work you're generating.

(optional) I put together a short playbook to get you started:
→ when to use Claude vs Claude Cowork vs Claude Code
→ what each of the 10 levels actually unlocks
→ the mental model for knowing which tool fits which task
→ the specific features that get you from 0 to 3     
→ the "letting go" exercises that separate 5 from 6
→ how to set up background agents (level 9)

Want it? you know the drill: 
1. 👍 like                        
2. 💬 comment "claude levels"              
3. 🔄 repost for good measure

Helping GTM Leaders & Founders Grow With GTM x AI | Fractional CxO | Building Linkedin Tools @ humanoidz.ai

everyone says AI agents need constant babysitting.

They haven't tried this…

The criticism is always the same: "sure it can start a task but you'll spend more time fixing its mistakes than doing it yourself." I used to agree. Agent v1 could work for 2 minutes before needing help. Agent v2 stretched to 20 minutes. Still felt like managing an intern who needed constant check-ins.

When Replit dropped Agent 3 a while back... I never had the issue. 200 minutes of autonomous work. Here's what changed>> it tests itself! When something breaks, it doesn't stop and ask what to do. It runs its own tests, identifies the failure, fixes it, and keeps going. The loop that used to require you sitting there watching... now happens while you're in another meeting.

In a project we were asked it to build a dashboard connecting to a client's Salesforce and Notion data. Walked away. Came back 3 hours later to a working app with auth, data sync, and export functionality.

In this weird world we're in now... the weirdest part is...: it builds other agents. You can have Agent 3 create specialized workflows that handle repetitive tasks across your business. Not just code. Any work that follows a pattern.

The mental shift:
→ old thinking: AI helps me code faster
→ new thinking: AI handles entire workstreams while I focus on strategy

What's still hard:
→ complex edge cases need human judgment
→ prompting well matters more than ever
→ you still need to know what you're building and why

You can do this too and get part of your life back (and enjoy your job again). Map out what you'd automate based on impact and ease. Start with high impact stuff that's straightforward to describe. I promise if you do this, you'll start liking your job way, way more.

stop sending meeting notes.

do this instead...

a demo is worth a thousand hours of explaining. I used to send detailed proposals to clients and lengthy briefs to teams = what we'd build and why. now we skip the explaining and just show something clickable. whether it's a prospect or a teammate.

here's a setup I use:
all my intake calls and notes live in a few places place. tldv, copilot, notion: whatever the source (depending on client's environment and compliance rules), it flows into the adapted system.

The magic is using's Replit's connectors to connect to all my notes and context. that means replit agents have context from every conversation. not just the last call... all of them. client history, past decisions, requirements we've already discussed (compliant data sharing)

after any meeting where work gets mentioned, i run an agent that: 
→ reads the relevant context from past conversations 
→ extracts what actually needs to be built (not just apps, can also be workflows or strategy docs) 
→ suggests what a first version could look like 
→ builds a working prototype
this works for new clients and existing projects. here are 2 scenarios:

1. new client? 
my follow-up mail or next call is never just a proposal. it's a proposal with a clickable poc.
Don't show up with a deck explaining what you'll do and why you're qualified. the client has to imagine what the end result might look like based on your words. they're buying a promise. i show up with something they can click through. they see the thinking. they see that i'm already working, not just pitching. and when the poc is slightly off? even better. they point at the screen and say "actually we need it to do this instead"... and now we're iterating on something real instead of debating abstractions. Your close rate will increase big time.

2. existing project? 
someone mentions "we need X for Y" in a sync? there's a working draft before the next meeting. someone has a good idea or a decision is made in a meeting, everyone agrees it's worth exploring, it goes on a list somewhere and often momentum disappears. the context is gone and nobody remembers why it felt urgent. now when something comes up: a workflow that needs fixing, a tool that would save time, a pilot worth testing, there's a working version to react to within 48 hours. the conversation shifts from "should we build this" to "here it is, what's wrong with it." (or is this even really a prio). ideas stay alive because there's something tangible to push forward

The mcp connections possible with Replit is what makes this sustainable.
replit isn't starting from zero. it knows the client. it knows what we discussed. the pocs are actually relevant instead of generic demos that miss the point.
i used to spend hours writing proposals explaining what i'd build.
now i spend that time letting them click through what i've already started.

what would change if every meeting that mentioned building something actually produced something?

we 100x'd our organic reach (SEO/GEO) without writing blogs

here's the exact playbook we used...

most teams are still obsessing over "keywords" and "clicks." meanwhile, we realized the real game has shifted to being the cited answer for the 800M+ users who skip Google entirely.

we call this GEO (Generative Engine Optimization)...and the basis of GEO is still SEO (two birds one stone) it turns out that 87% of what AI cites already ranks on Google, but simply adding specific "GEO" tactics unlocks up to 40% more visibility. (the secret is stopping the "content treadmill" and structuring what you already have so LLMs can easily extract the facts)

anyway, i put together a full playbook covering: 
→ the technical foundation (why Bing Webmaster Tools is suddenly critical) 
→ the "Best X" list strategy that captures 43% of AI citations 
→ how to write "Quotable Sentences" that LLMs love to steal 
→ the Schema Markup trick to feed data directly to the bots 
→ the "Freshness" signal that keeps you in the answer box 
→ the mistake of ignoring "zero-click" searches 

turns a standard website into an authority engine for both Google and AI.

want it? you know the drill:

1. 👍 like
2. 💬 comment "seo/geo playbook"
3. 🔄 repost just to be sure

(make 100% sure we're connected) will send it over.

Thanks to the GenAI ⚫️ Circle team for compiling this massive research. Special thanks to the beautiful guidance of Walid Boulanouar, Jan-Willem Bobbink and Roald Larsen

The Ultimate SEO and GEO Playbook 2026

This is probably the biggest news you missed (it's huge)

and it happened just yesterday...

Apple and Google announced a multi-year deal. Gemini will power the next generation of Apple Intelligence. Not as an app. As the foundation for Siri and Apple's on-device AI. Why? Coz it didn't work with OpenAI (apple Intelligence sucked)... Pls read that again. Google's models are becoming the substrate beneath a billion Apple devices. Think about it... Google created Android. Now Google's AI will be the brain inside iOS. Crazily ironic. Gemini won't be something you open. It will be something that's always there, watching your calendar, reading your emails, understanding your files. Context you never have to provide because the OS already has it. This is invisible AI, this is AI at the OS level.  We've been hacking context together for years now. Copy this document. Paste it into Claude. Upload that screenshot. Drag in the PDF. It works. But it's clunky. You are the middleware between your information and your AI.

That era is ending.

The real AI race was never "which chatbot wins." It was always "whose model becomes the operating system layer." The place where AI doesn't wait for you to summon it. Where it sees what you see, automatically. Apple just answered that question for their entire ecosystem. And they answered it with Google. Again the irony...Google created Android. For 15 years, Apple and Google have been the two poles of the smartphone wars. Now Google's AI will run inside every iPhone. It happened because Apple Intelligence wasn't working. The OpenAI integration underwhelmed. Third-party apps like Perplexity were handling basic tasks better than Siri. Apple looked at the gap between their ambitions and their capabilities, and made a pragmatic choice.
Pride is expensive. Apparently more expensive than a partnership with your oldest rival.

At the end of last year, we shared a prediction with our GenAI ⚫️ Circle. AI would move to the OS layer. Manual context management would disappear. Google already had multiple teams working on this. We expected it within a few months or years. It took a few weeks. The mechanism surprised me. I expected new operating systems built from scratch. Instead, Google's AI is colonizing the most successful existing OS on the planet. If you're building AI apps that depend on users manually feeding them context, your moat just got thinner. The OS will do that job soon. For free. By default. The window for "AI wrapper" businesses is closing faster than most people realize.....

Thank you Andreia Cândido Saraiva for sharing the news with us this morning!

Our 2025 AI predictions were almost spot on

Here's what's gonna happen in 2026...

Every year, our community of 400+ heavy AI adopters gets together to call what's coming. No hype. No hot takes for clicks. Just practitioners watching the field closely. Last year we nailed 5 out of 10. We correctly called agentic breakthroughs, GenUI, the rise of a major Asian competitor, and AI avatars becoming indistinguishable from humans. We missed on LLMs taking 50% of search and on small models overtaking large ones.

Pattern we noticed: we're good at predicting technology. Less good at predicting market dynamics.

So what's coming in 2026?
The short version: AI stops being a feature and becomes infrastructure. We'll see 2 to 3 AI native operating systems emerge where models see everything you do automatically. No more hacking context together through browsers and file uploads. Google already has three teams on this.

Companies will start managing AI staff members alongside human workers. Actual roles, responsibilities, performance metrics. The "copilot" framing is already outdated. Open source will eat real market share. DeepSeek matched o1 for $5.6M and wiped $600B off Nvidia in a day. That cost advantage is too big to ignore.

Two internets will emerge. One for humans with interfaces and clicks. One for agents with just APIs and transactions. No UI needed. Just outcomes.
At least one legacy SaaS giant gets disrupted by an AI first startup. At least one CEO gets publicly fired for failing to implement AI strategy. And a major AI powered cybercrime breach will trigger the first serious wave of regulation. There will also be real backlash. Not from Luddites. From workers and consumers angry about companies using AI as cover for mass layoffs.
And the leading models at the end of 2026? Probably don't exist yet. A year ago, who had heard of DeepSeek?

Full breakdown in the carousel. Save it. Share it pls :) 

Check back in December.

Now it's your turn! What are we missing? What are your predictions? The bolder the better of course

2026 AI Predictions by GenAI Circle 

This is the most important skill to teach your kids

It isn’t what you think…

(This also applies to us parents). 
As parents we want to give our child an advantage. But we/you might be optimizing for the wrong things. Children arrive curious in this world. They disassemble toys. They ask uncomfortable questions. They have an almost scientific relationship with the world, poking at it to see what happens. Then we educate them.

Education, as currently constructed, is a basically a compliance training program. Sit in rows...wait forinstructions. Measure yourself against rubrics designed by strangers. The entire apparatus optimizes for manageability, because manageability scales. You cannot coordinate thirty children if each one is running independent experiments. We ask our kids to chase grades. Some of us treat childhood like a checklist. But the checklist era has ending. And most of us are preparing kids for jobs that will either not exist or not matter.

So what is the skill?

It isn't technical, it isn't "learning to code" or "understanding AI" or whatever panic is trending this quarter. It's also not just adaptability or growth mindset.

It's self-direction.

Self-direction is the capacity to move without a map. To decide what matters and pursue it before anyone gives you permission. To sit in ambiguity without freezing. To treat setbacks as feedback rather than judgment. Most adults cannot do it. Strip away their job title, their manager, their performance reviews, their calendar invites and they become paralyzed. Not from lack of talent. From lack of practice. They spent two decades learning to follow curricula and then wonder why they can't write their own. AI will commoditize most knowledge work within a decade (less). Anything reducible to process will be automated or augmented beyond recognition. What remains valuable will be precisely what machines cannot replicate. Machines cannot want things. They cannot choose a direction based on conviction and taste. They cannot decide that this particular problem is worth years of their life, then actually commit those years.

Self-direction is not chaos. It is not "follow your passion". It is a specific relationship with uncertainty. Self-directed people treat problems as puzzles, form theories about what might work, test those theories. When results disappoint, they conclude that particular approach was wrong and they generate another one. Here's what makes this valuable. Almost no one will do it. Self-direction is illegible. It produces kids who are uncomfortable to be around because they don't automatically defer to authority. Most parents will keep optimizing for grades and credentials because that's what they can measure and brag about. Which means the few who build self-directed kids will produce adults who set the terms while everyone else waits for instructions that aren't coming.

How? Introduce real stakes, stop rewarding outcomes only... and get out of the way more ;)

while everyone kicks off 2026 vibecoding a billion dollar app

I built something much more important...

A donkey shed. 

About a month ago I shared how I used Gemini's Nano Banana to generate an accurate blueprint for this build. A bunch of week-ends later we're well underway. My arms as the human in the loop. Made some modifications along the way. Still at MVP stage, but pushing to production within the next few weeks 🫏

Here's what strikes me. We keep measuring AI intelligence by how well it handles language. Can it write essays? Can it reason through word problems? Can it chat like a smart person? But watch what Veo 3 does with water and lighting and complex physics. Watch what Gemini does when you ask it to think spatially about load-bearing walls and roof pitch. These systems are developing a kind of world-model intelligence that has nothing to do with verbal fluency. Maybe we've been holding the ruler backwards?? Superior artificial intelligence might not look like a better conversationalist. It might look like something that understands how physical reality actually works. Protein folding. Fluid dynamics. Structural engineering. Domains where formal reasoning already exceeds human capacity, but we don't count it because the machine can't narrate the experience. How human-centric of us...

Artificial Donkey-shed Intelligence is already here.

Anyway. Back to cutting wood tonight.

The happiness paradox is hidden in your phone 

You’ll never guess where…

In 2010, two Harvard researchers built something unusual. An iPhone app that pinged 5000 people at random moments throughout their day. It asked two simple questions. What are you doing? What are you thinking about? The finding was almost too clean to believe. People’s minds wandered 46.9% of the time. Nearly half their waking hours spent somewhere other than the present moment. And here’s what mattered. Mind wandering was the cause of unhappiness, not the consequence. The nature of the activity barely moved the needle. Whether people were working, commuting, or making love, what predicted their happiness wasn’t what they were doing. It was whether their attention was actually there.
This research was in 2010. Before the attention economy had fully colonized our lives. Before the average person touched their phone 2617 times per day. Before “doom scrolling” entered the lexicon. ~15 years later, we’ve built an entire economy around manufacturing mind wandering. Every notification, every infinite scroll, every autoplay video is an invitation to be somewhere other than where you are. The business model of the internet is, quite literally, the monetization of attention. Killingsworth and Gilbert revealed something more fundamental. Attention isn’t just an economic input. It’s the substrate of human wellbeing itself. Pascal wrote in the 1600s that all of humanity’s problems stem from our inability to sit quietly in a room alone.

Now, if you’ll excuse me, I’m gonna go back to scrolling LinkedIn.

No alternative text description for this image

Claude vs gemini vs copilot vs chatgpt.

we're watching the wrong race. here's why...

Almost nobody is talking about this yet. That's about to change. 

What if the real competition isn't between labs? What if it's between approaches? LLMs vs EBMs. Language models vs Energy-Based Models. Yann LeCun spent years on an island (bit dramatic but kinda true), waving a red flag at the scaling crowd. He argued that predicting the next token would never get us to reasoning. That intelligence required world models, not pattern completion. That a fundamentally different architecture was needed.
The field mostly ignored him. Or mocked him (check X). Then Ilya went on Dwarkesh's podcast and said something crucial was missing from the current paradigm. Suddenly everyone agreed with what LeCun had been saying for half a decade.

Now LeCun has stopped just talking about it.

Logical Intelligence launched this week with LeCun as chair of their research board. Their first model, Kona 1.0, is built on Energy-Based Models EDMs.
The demo is Sudoku. Which sounds trivial until you understand what it's actually testing. Sudoku isn't about knowing facts. It's about satisfying constraints simultaneously. Every row, column, and box must cohere. You can't solve it by predicting what number "usually comes next." You solve it by finding the configuration where nothing contradicts anything else. This is exactly what LLMs struggle with. They generate tokens sequentially. Each choice is locally plausible but globally they drift into inconsistency.

EBMs work differently. They don't ask "what token is most likely?" They ask "what configuration has the lowest contradiction?" Low energy means coherent. High energy means impossible. The system searches for states where everything fits together.

The Kona demo apparently beats GPT and Claude on Sudoku consistently. Not because Kona knows more. Because the architecture is designed for constraint satisfaction rather than sequence prediction. You can check the demo yourselves. Whether this generalizes beyond Sudoku is the real question. LeCun's bet is that reasoning itself is constraint satisfaction. That understanding means having an internal model where valid interpretations settle into stable states and contradictions get rejected.

The transformer crowd bet that scale would produce reasoning as an emergent property. That bet hasn't fully paid off. Chain-of-thought helps. But it's a patch on an architecture that wasn't designed for this.

The most interesting development isn't another GPT version beating another Claude version on another benchmark. It's whether a completely different paradigm can do what scaling couldn't. Forget OpenAI vs Anthropic vs Google. The real question is whether we've been building on the wrong foundation entirely.

the best ROE (return on energy) to focus on atm...

90% of people are ignoring at least half of these...

- build your name/audience (audience + distribution = the real moat now)
- Claude Code (to manage everything)...or CoWork if you hate the terminal
- MCP as the integration layer: connecting AI to everything (notion, canva, meta ads your CRM). this is where the leverage is. keep human in the loop.
- human psychology as moat... when everything else can be automated, the stuff that can't is what matters
- Community (from a simple whatsapp group to a full on membership group) can't be replicated, filters out the ai-slop and unnecessary noise of SoMe networks
- Learn to "stop the scroll". The attention economy is the most aggressive it's ever been. 
- pSEO works (to rank on SEO and GEO and attract/convert customers)
- GEO > SEO: LLMs are stupidly sensitive to keyword stuffing. papers keep proving it. optimize for AI answers not just google
- self-direction (rules are chaning monthly, learn to navigate without guidance)
- RAG RAG RAG (the repo is the new org chart, he who controls the repo/the context is the true CEO)
- Founder-led + Leadership-led content (this is the way)
- feedback loops = systems that feed success data back into themselves. that's when scaling gets hands-off
- the awareness gap... most professionals still don't know half of this tech exists (claude code, mcp etc). the window is still open
- now your turn...   [insert your thing]

I feel like a Eurostat researcher now. Thx Claude Code.

Was wondering where USA earns "best country in the world" status versus Europe. So I pulled the data.

The answer is cleaner than the X/Twitter debates suggest. America dominates on power and wealth. Number one military. Highest soft power score at 78.8. GDP per capita of $85,000 versus the EU average of $54k. Median disposable income of $42,8k versus $23000. And 50% of adults with college degrees versus 42% in Europe. On raw economic and geopolitical muscle, America wins and it's not close. But that's where American dominance ends.

Europeans live longer. The Swiss hit 84.4 years. Americans hit 79.5. Infant mortality in America is 5.1 per thousand. In Sweden it's 2.3. America spends 16.7% of GDP on healthcare. The Netherlands spends 10% and gets better outcomes. The work gap is pretty crazy. Americans average 35.8 hours per week. Germans average 26.5. Over 10% of American workers pull 50+ hour weeks. In the Netherlands that number is 0.4%. America mandates zero paid vacation days. Sweden mandates 38. The happiness scores land exactly where you'd expect. America at 6.73. Sweden at 7.34. Netherlands at 7.32.

France has baguettes 🥖 and croissants 🥐 (so we win across the board).

TL:DR... America generates more wealth and more power. Europe generates more health and more happiness.

Crazy I could pull this up in a few minutes.

Now go ahead and tear me apart in the comments for bad data or flawed methodology :P

PS: Will also break down USA into some individual states and share in the comments in a bit

We were told we can't have fast, cheap AND good

I'm gonna p*$$ a lot of ppl off with this...

For decades I've been told... no David you can't have cheap, fast and high quality. Pick 2. Project management, product, business in general has operated under this kind of holy trinity. The Iron Triangle (which is almost never represented as a triangle anyway). Every internal meeting, agency pitch, every vendor negotiation, every internal resource battle assumed this tradeoff was the ultimate argument. Want quality? Pay more or wait longer. Need it tomorrow? Sacrifice one of the others.

I'm calling it dead. Not for all use cases of course. But for many many many use cases.

Last month we ran visual asset production for two clients. Full ad creative. The kind of work that used to require a designer, a couple rounds of revision and a week of calendar time. We built production ready assets at scale. Quality indistinguishable from our old process. The ads are converting. Retargeting and cold traffic. First click acquisition. We put bold claims in the creative, the kind that actually "stop the scroll" and they're driving to quality product pages. This isn't a proof of concept anymore... it's live and generating revenue. I have more and more of these examples.

So I'm calling the Golden Triangle dead. Those still applying it are late to the game.

Now of course the real question that should keep us up at night... When everyone can produce good, fast and cheap, then good, fast, and cheap becomes noise. Not signal. The advantage disappears the moment it becomes universal. I wrote recently about signaling theory. When something is free to produce, economists call it cheap talk. Cheap talk is worthless. The peacock's tail works as a signal precisely because it's expensive. The cost is the proof.
So when production costs collapse, what's left? 

There's still an arbitrage possible while most play catch up. But before you know it (couple years max), the only things that will remain scarce = Taste +  Judgment + things AI can't produce at scale (referrals, human interactions,closed communities etc) and/or the strategic thinking that decides what to make and why + the willingness to commit to a point of view for example.

In an age where distribution is “everything”

LinkedIn is B2B-unavoidable. And here’s why….

In this Semrush data from end 2025, LinkedIn is second. not Wikipedia. not news sites. LinkedIn.
and honestly this makes total sense when you think about it. if you’re building b2b tools, where else are llms gonna find people talking about their actual work experiences? product reviews, industry takes, career stuff…it’s all on LinkedIn. LLMs don’t privilege prestige. They privilege density of genuine human perspective. Reddit wins because real people argue about real problems in their own words. LinkedIn wins because professionals share actual experience from actual work. 

Which means your LinkedIn content isn’t just reaching your followers anymore. it’s training the models. showing up in ai-generated answers. becoming part of how these systems understand your industry. Being cited by an LLM is becoming what ranking on Google was ten years ago. But you can’t game this the way you gamed SEO. Keyword stuffing doesn’t work when an AI evaluates semantic content. Link farms are useless when relevance is measured by actual usefulness. 

anyway, Walid Boulanouar and i spent a lot of time figuring out how to grow through genuinely great content on LinkedIn. what the algorithm rewards, what gets reach, what converts. and we baked all of it into ViralBrain [.] ai

the game really is different now. your content has two audiences: humans scrolling their feed…and machines deciding what to cite.​​​​​​​​​​​​​​​​Again, we’ve entered an era where audience and distribution are the hardest and the most critical asset.

Thanks to Andréa Bensaid for sharing the chart today!!

the AGI race isn't about intelligence anymore.

it's about loop closure.

and the loop is clearly about to snap shut. Demis Hassabis (Google) and Dario Amodei (Anthropic) are looking at the same map. they just disagree on the speed of the car. the frame: “AGI” isn’t a chatbot that feels human but rather a system that can do most cognitive tasks at an elite researcher level, across domains. the hinge is the self-improvement loop. models are already good at coding. once they can do end-to-end engineering and research (not just snippets), they start writing their own next-gen code. Just check out Ralph Wiggum loops….apps are already being brute forced today by ppl I know. that’s why one camp says 2026–2027. the more cautious camp says “50% by end of decade,” because some breakthroughs still need physical experiments and the weird part of science: choosing the right questions.

The labor market is entering a 5-year 'efficiency squeeze.' not a cliff, but a 15% YoY reduction in junior headcount demand…. entry-level white-collar work is the first domino. estimates are as high as 50% of junior roles disrupted in 1–5 years, with the earliest signal being hiring slowdowns (interns, new grads, coordinators). the advice isn’t “learn to code.” it’s “become unbelievably proficient at using AI tools” so you can do senior-shaped output early. then it gets existential. if we actually approach post-scarcity, the hard problem shifts from money to meaning. people don’t just need income. they need a reason to wake up.
 and the technical bet: mechanistic interpretability (actually looking inside the model’s “brain”) to understand why deception or power-seeking behaviors emerge. dual-use is the curse: the same system that accelerates science (think AlphaFold-style wins) can accelerate harm.
the playbook for the next 12 months:

→ track: can models do end-to-end engineering without humans closing tickets
→ measure: “loop closure” (model helps build the next model)
→ expect: coding + math automate first (they’re verifiable)
→ wait for: robotics + world models breakout moments
→ plan for: junior roles shrinking, “AI-native” roles exploding

what to do now:
1. audit your org for “verifiable work” (anything with tests/checks)
2. redesign entry-level paths around AI copilots, not busywork
3. treat compute/chips as a safety lever, not just a market commodity

But mostly: think AI-first. Start from scratch….what would this business look like if it was born/built today?

TL:DR… the AGI race is a loop-closure race. jobs and geopolitics move faster the moment the loop closes. most people will read this and do nothing.
PS: if your strategy is “we’ll adapt when it’s obvious,” you’re already late. Don’t believe me?? Just onboard yourself to Claude Code. You’ll see what the future looks/feels like

c’mon…

admit it…

this is getting too real

taplio

The Battle

Dale Thomas

David Arnoux

Followers

Posts

Engagements

Likes

Performance ScoreDavid Arnoux's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Dale Thomas

David Arnoux

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}David Arnoux's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreDavid Arnoux's LinkedIn Performance