Jeff Frick

I am not telling you something to sell you something. I am sharing something to entertain, educate, and get you to stop and think for a minute. | Top Voice | Friday Five | Turn the Lens | Work 20XX

With over 30 years of experience in the media industry, I am a podcast host and a media entrepreneur who is passionate about the future of work, leadership, and innovation. I have founded and led Menlo Creek Media, a company that helps leaders, marketers, and communications professionals navigate and leverage an AI-driven asynchronous world, by offering executive training for podcast hosts and guests, content creation and media library curation and amplification programs, and community development programs.

I also host two podcasts, Work 20XX and Turn the Lens, where I facilitate the spread of best practices, insights, learning, and wisdom on topics that matter to the modern workforce and society. I aim to be a trusted source of information and inspiration for my listeners, by featuring smart and diverse guests, engaging in lively and thoughtful conversations, and sharing nuggets of knowledge. My goal is to help people get more impact with their communications and content investments and to empower them to thrive in a time of rapid change and uncertainty.

Founder & Principal

Please visit www.MenloCreek.com for a more comprehensive list of services and resources. 

Helping Leaders, Marketers, and Communications professionals navigate and leverage an AI driver asynchronous world. 

Host of 'Work 20XX'  with Jeff Frick show on the Future of Work 
Host of 'Turn the Lens' with Jeff Frick on Leadership and of other topics of interest.

We are able to take a model that is trained on very general purpose physical interaction data, and we train on more and more of it. And every single task that we are tracking continues to get better. - Pete Florence

To be honest, it doesn't even really necessarily matter too much what the x axis is. Other than like the x axis is something that we can continually do more of and the y axis is some measure of how good the robots are. And we see that we can just continue to pour in more and more effort on that x axis and the y axis continues to get better and better. - Pete Florence

And that is like a landmark type of moment - Pete Florence
----
Landmark moment.

I was blown away when I saw the slide earlier in Pete's keynote. Not enough parameters, the models crap out (a technical term), but push past 7 billion, and the magic happens. Improve on one, improve on all. Pete and the team at Generalist recently announced a 10 billion parameters plus foundational model.

So excited to share this conversation with Pete Florence, Co-Founder and CEO of Generalist. Generalist is another example of the latest round of startups around 'Embodied AI', AI that does things in the environment, among us. Pete has a deep research foundation, and we retraced some of the key moments in getting from the highly specialized world of industrial robots, to the 'every use case is a corner case' of interacting in the physical world outside a pen, to respond to any number of task demands and unforeseen environmental challenges standing between the present, and completing the assigned task. 

As Pete said, take the model and make it the robot brain.
Make it big enough, and the time you spend training on one thing, improves the performance on all things... 
Bang. Stops you in your tracks. 

Thrilled to co-release my conversation with Pete Florence on Turn the Lens, in collaboration with Humanoids Summit, brought to you by ALM Ventures, from Humanoids Summit 2025 Silicon Valley, December 2025.

Full Length Conversation
Pete Florence: Generalist, Scaling Laws, Train One Improve All | Turn the Lens with Jeff Frick Ep46 
https://lnkd.in/gta8S7Da

In collaboration with Humanoids Summit, brought to you by ALM Ventures, 
from Humanoids Summit 2025 Silicon Valley, December 2025
Thanks Modar Alaoui and Jesica Chavez 🤖🎙️ Chavez 

For more on Humanoids Summit, including 
May 2026 Summit in Tokyo visit 
https://lnkd.in/ga487tGq 

For more from Humanoids Summit SV 
https://lnkd.in/geeUa5Dp  

#embodiedai #physicalai #humanoidssummit #turnthelens #robotics #humanoids #Generalist

• Embodied • Scaling • Policy • Dexterity • Trades • 

First Friday Five of 2026. Embodied AI, when AI has arms and legs and can do things. When every case is a corner case, never before experience in training. (for instance, EVERY Waymo ride)

Thrilled to co-release the first five interviews we did at Humanoids Summit SV in December in collaboration with Humanoids Summit, ALM Ventures, Modar Alaoui and team.

And so much more to come. 
#Embodied #Scaling #Policy #Dexterity #Trades #FridayFive #Newsletter #TurnTheLens #Work20XX

It's been a few weeks, but we're back in the saddle. Closed out last year with ten interviews from at the Humanoids Summit SV in Mountain View, CA at 

Open article: • Embodied • Scaling • Policy • Dexterity • Trades • by Jeff Frick • 5 min read

• Embodied • Scaling • Policy • Dexterity • Trades •

Business Data Products | Data Product Storytelling | Data Product Pyramid | Data Product Marketplace

Ask me about delivering e2e AI native business processes using the Nebulyx AI platform and our A.D.O.P.T. framework - AI + Data Object Graphs and Data Product Thinking. Using our Data Product Pyramid framework on our AI Digital Twin solution and generative AI, AI Agents, and advanced analytics with Data Object Graphs to deliver the “Composable Enterprise”: a low friction, adaptive eco-system of re-configurable data products that runs at the speed of the business.

This includes 
1- Using LLMs/AI to create advanced analytical and Data Products in reconfigurable Data Object Graph based eco-system
2- Creating a Lingua-Franca that allows business teams to communicate analytics asks with data and tech teams.
3- Creating an Adaptive Business Strategy - The business strategy can start small and evolve and adapt as the business changes.
4- Implementing as a rapid delivery, highly valuable data and analytic use-cases as Data Products
5- Creating A portfolio of actual and specific real world Data Products that can used and re-used across the business.
6- Creating a Data Product Marketplace where internal and external customers can create, prototype, search and use data products, including: Datasets, ML Models, BI Dashboards, Graph analytics
7- Creating hybrid cloud polyglot data infrastructure to deploy Data Products

Jon Cooke

Context Graphs: A Time Machine for Business Decisions ⏳

I do think there is a real opportunity with context graphs (so def agree with Foundation Capital on their concept - link in comments)

We’ve had databases as systems of record for decades.

But they only record the transactions, the what, but not the why.

Agents create many dynamic decisions to get their goal, with context graphs being historical record of the decision process.

Not just what happened, but why the Agent made the decision.

Every run of the Agent can produce a very different graph.

All the relevant information on each node.

A snapshotted, record of the decision states, transitions with applicable at the time, knowledge, constraints + other information all preserved in queryable form.

It’s like a time machine for a business. 🕰️

“How did we react to that market shift in Q2 2024?”

“What did we know when we made that pricing decision?”

“Why did we approve that exception last March?”

Not reconstructed from databases, not pieced together from emails.

Actually captured, in context at the moment of decision.

Here’s where it gets interesting. 🧩

Each decision path (specific graph) is potentially a different shape.

Different number of nodes and edges and different data on each node.

Bridging these branches in a queryable way becomes a fascinating problem.

How do you query across thousands of decision graphs, each with different structures?

Our approach: 🔄

Data Object Graphs with a specialised LLM that has temporality encoded as a first-class citizen.

It can dynamically consume many “graphs in time” without needing uniform structure.

These graphs persisted enabling querying across decision history to find patterns and learn from the past.

Not just what the business did, but how it thought, reacted, learned and how it evolved.

For me that’s a real promise of this type system of record from Agents that use graphs.

Graph based approaches like this are the next generation of reliable, enterprise grade agents.

Shout out to Ashish Verma who did a great post on temporality (link in comments).

My macro view (across many decisions over time) dovetails nicely with his excellent micro view (inside one decision).

Nebulyx AI

Preview your AI native business with Nebulyx AI (AI Digital Twins) and A.D.O.P.T. - AI, Data Object Graph and Product thinking | Data Product Workshop Podcast host

Context Graphs: Time Machine For Business Decisions

Context Graphs and Knowledge Flywheel for Agents🐕
 
Context graphs are "so hot right now". Seeing (& participating in) loads of great discussions.🔥
 
From these, two operating modes for knowledge have surfaced:
 
1️⃣Traditional Knowledge Management: 
People in a room interviewing, reviewing, discussing, then creating structured knowledge artefacts KGs etc.
 
2️⃣Dynamic Knowledge Assembly (Agentic mode)
Agents executing from a core question (either human or machine initiated) dynamic chains of reasoning, assembling knowledge for each reasoning step, deciding on the next step and repeating. 
 
This is where context graphs fit.
 
Dynamic Knowledge Assembly is a mouthful, hence the "Knowledge Flywheel" 
 
A living blueprint of a business, that changes at the speed it does.
 
The key shifts are: design-time to runtime, managed to dynamically assembled.⚡
 
Three Knowledge Layers📚
Understanding what type you're working with matters.
 
🌐Domain: Regulations, industry manuals, clinical guidelines. External knowledge that governs your field.
 
📋Explicit: Company policies, process flows, transactions. Internal knowledge that's written down or captured in systems.
 
🧠Tacit: Undocumented expertise. Overrides, edge cases, judgement calls. Knowledge that only surfaces when decisions are made.
 
Traditional approaches try to capture all three upfront manually. The comments across various threads show how hard it is and how long it can take. ⏳
 
The Knowledge Flywheel 🔄
Our AgentDOG approach is different & operates in full Agentic mode (2).
 
I.e. we don't try to capture everything upfront, we let the flywheel turn.
 
Populate → Execute → Capture → Learn → Encode → Execute...
 
🔹Populate: AI-native semantics extracts meaning from domain docs and explicit knowledge. 80% done automatically. SMEs verify & tune the last 20%.
 
For our Doctor Dog example 🩺, we fed in NICE clinical guidelines (41 pages). 96% vocabulary coverage. 80% critical questions answered out of the box. SMEs then tune the 20%.
 
🔹Execute: Agent navigates the decision space. States, transitions, constraints as nodes and edges. 
 
Context graphs persisted as decision traces, on completion.
 
🔹Capture: Tacit knowledge enters here. SME overrides a decision? Captured with full context: state, reasoning, constraints. Knowledge born through action. 👤
 
Also AI Digital Twins can run scenarios -> SMEs override real situations, with capture. 
 
Run-time not design-time thinking
 
🔹Learn: Reinforcement from outcomes. Patterns emerge & precedent builds. 📈
 
🔹Encode: Knowledge crystallises back with new states & transitions. 
 
Flywheel gets smarter. ✨ Then it turns again.
 
Why This Works ✅
- Tacit knowledge is easier to capture at runtime than upfront.
- Full context available in the moment beats interviewing experts weeks later.
- Context Graphs govern/capture action.
- Learn, evolve from and through actions.
 
The flywheel turns. The knowledge lives. 💡
 
Nebulyx AI

Context graphs are about agents making and learning from their decisions. 

There seems to be a lot of confusion as other domains and problem spaces are being conflated with this concept. 

As Jaya Gupta states, they are "...decision traces are captured in the execution path, not defined upfront in workshops, not reconstructed after the fact via ETL. They emerge as a byproduct of agents doing work." 

The article "AI’s trillion-dollar opportunity: Context graphs" by Jaya Gupta and Ashu Garg (Link in the comments) emphasizes that this discussion is about agentic decision-making, not human decision-making. 

Their perspective is clear: "AI agents are hitting a wall. The wall isn't missing data. It's missing decision traces." 

Context graphs serve as the runtime artefact for agents, with the "context" term coming from the typical information passed into each LLM inference/query during execution which the AI world calls "context"  

Although worth noting context graphs contain more than just information passed i.e. all the foundational information the Agent uses to make it's decision.

They provide a decision log/trace generated through purely agentic execution, rather than extracting data from humans post-factum. 

The same agents and other agents can then learn from these specific agentic decisions. 

The graphs are in the purview for companies deploying agents today. 

They are are distinct from: 
1. How people make decisions, including capturing those decisions. 
2. Determining when to use agents versus people or other methods. 
3. Deciding whether to model workflows upfront or adopt agentic approaches (context graphs fall into the latter). 
4. General knowledge management. 
5. General metadata approaches. 
6. Typical Knowledge graphs - which typically try to capture meaning up front 
7. The best graph modelling approaches (eg RDF etc).

While these are all important questions, and definitely part of a bigger question of how AI and Agents are changing these, it's crucial to maintain clarity and not blur the boundaries of the problem space.

With our AgentDOG framework, in our Nebulyx AI platform, we use graphs, extensively for the core execution and for learning and we persist them automatically (as Data Object Graphs).

I.e. use them for run-time execution and persistence -> the "institutional memory", as the article puts it.

I know there are a lot of questions around this specific topic, and indeed the others I've listed. We've been building agents at Nebulyx AI, using graph based for execution, and navigating these exact decision points with clients. 

If your org needs clarity on your agent strategy, decisioning approach, and knowledge management, these can't be tackled in isolation. 

These decisions need to be holistic and consider agentic design/execution/appropriateness as in conjunction with the other topics.

Which requires a broad perspective. Which we have and we are happy to help you navigate this complex but crucial space.

Context Graphs: Agents Making and Learning from THEIR decisions

AI Native Semantics: Move 37 For Encoding Knowledge

Saturday reflections (& prob get me in trouble)🤔

Move 37. AlphaGo’s legendary move against Lee Sedol was a move no human would make. 

Looked wrong to every expert but won the game.

Had my own Move 37 moment recently.

The problem
Extract semantics from regulatory domains and apply to specific companies. 

Different companies, different contexts, same underlying domain.

Classic ontology problem, right? So I tried the ontology first approach.

But kept breaking.

Not because methods were bad. 

Because, I know I’m going to get pushed into debates about modelling (which it isn’t) they’ seemed more designed to build libraries, not decisioning systems.

Ie there’s an issue with the abstraction approach I kept hitting.

Real example:

A regulation has multiple versions. 
Amendments, additions which company specific legal interpretations are then layered on top.

These compound. ie are relevant at the time of that version. 

So understanding, intent, decision, and action are rooted in the specific state at that relevant moment.

Classic ontology tries to abstract this into universal views with cross temporal entities and stable relationships.

But the point of the regulation isn’t the abstraction. For an org it’s the key decision it needs right now, for this company, under these interpretations.

So there’s tension there, in objective, which I’ve run slap bang into.

So I went back to basics. Our Knowledge Flywheel doesn’t abstract meaning for humans to interpret later. 

It encodes domain truth directly. 

Regulations, guidelines, policies and decisions, all versioned & bounded at the specific time.

I created a new LLM model architecture to solve this tension. Not probabilistic, deterministic ie question, same answer with full provenance.

Information & semantics encoded directly in the model ie no upfront human translation layer & no information loss.

With our AgentDOG, decisioning system, kept coming back to this:

Measurement, intent, decision, action - they’re contextual and specific to the process & specific to the point in time.

Ie Can’t separate intent from execution in practice.

Move 37
This felt wrong at first. Every instinct said: you need the ontology. You need humans to define meaning first.

But the results said otherwise. 80%  automated upfront without human intervention, deterministic & grounded.

Humans absolutely need to review and adjust (as not 100%, but close) + control & override when wrong.

Also we still need to understand & encode semantics but the assumption that meaning must flow through human abstraction before machines can use it, may no longer hold.

We are in a new era of AI-native semantics with meaning encoded directly in an AI
native way and rooted in the specific, not abstracted into the general.

I’m still on this journey and lots to learn and def standing on the shoulders of giants, but the results so far are very encouraging.

Happy Saturday.

Nebulyx AI

AI-Native Semantics: Building the 21st Century Organisation 🏢

Last century playbook: spend years & millions defining upfront, try document every process, map every workflow. 

Then pray the business doesn't change before you're done.

Spoiler: it always does.

⚡The 21st Century Organisation is nimble, reactive & adaptive.

Not because they planned better. Because they stopped trying to plan everything.

Business moves fast (esp. these days):
- Time-based offers launch & expire
- Seasonal fluctuations shift demand
- Customer feedback rewrites priorities
- Campaign results reshape profiles
- Regulations update 
- Competitors pivot

The organisations that win are the ones who can respond to it, fast.

This means living Agents and Knowledge across the organisation with highly skilled humans using Agents to accelerate & doing the grunt work

Agents spin up as needed. Sales agents. Support agents. Compliance agents. Product agents. 

Each one drawing from the Knowledge Flywheel (Context Graphs at core). 

Each one building own state machine at runtime.

Not predefined workflows, dynamic navigation through decision space.

Powered by Knowledge Flywheel 🔄

🌐Overarching Layer: Domain knowledge (regulations, industry standards) available to all agents 

📋End-to-end explicit knowledge (company-wide policies, cross-functional processes)

📋Business Function Layer:  Local explicit knowledge (team processes, function-specific rules) 

🧠Local tacit knowledge (judgement calls, overrides, expertise from function)

Agents don't start from scratch, they inherit from the flywheel. 

Domain plus E2E plus department-specific. Then they execute, capture, learn, encode, repeat.

Flywheel turns at every level & the organisation learns as a whole.

Humans at Strategic Points 👤

People relocate to where they matter most:
🎯 Customer-facing moments: Where relationships win or lose, and empathy counts.
🎯 Core product/USP: Where differentiation lives, & creativity drives advantage.
🎯 Override points: Where edge cases surface, where tacit knowledge is born.

Humans teach the agents & agents handle the scale.

Time-based offer launches tomorrow? 
✅Spin up agents, flywheel has the domain & company knowledge, feed in the campaign specifics, go.

Seasonal spike hits? 
✅Agent pool scales as knowledge already encoded.

Customer feedback shifts strategy?
✅Update the explicit layer, agents adapt, flywheel turns.

Ashish Verma's awesome post, Context Graphs: Temporal Reasoning highlights this (link in comments) 

No 18-month transformation programmes. No outdated process docs. No "need to re-document everything."

The Contrast ⏳
Old way: Years defining. Millions spent. Brittle artefacts. Outdated on delivery.

New way: Knowledge Flywheel turning. Agents spinning up. Humans at strategic points. Adapting business speed.

The 21st century organisation doesn't try to know everything upfront.

It builds the flywheel, positions the people and lets the agents run.

Nebulyx AI

AI-Native Semantics: Building the 21st Century Organisation

Dynamic Emergent Systems: The Legacy Thinking Holding Us Back

Context Graph conversations have surfaced a key tension: "How do we deal with change?"

How many massive migrations (data and system) do we do every year because systems can't cope with change?

I worked on a system with a clever SME, defined all the flows upfront, UAT went well. But when we ran it in production it fundamentally failed.

"Everybody has a plan until they get punched in the face" - Mike Tyson 🥊

Mark Stouse put it perfectly:
"We've been taught that if we can simply make processes stable and documented, it's all good. But what we control is only about 70-80% of why and how X happens, and when the uncontrolled portion also is volatile and rapidly changing, the process rapidly loses relevance."

That uncontrolled 20-30%? That's where reality lives. And you can't capture it upfront, it continually emerges.

The assumption holding us back 🚫

It's amazing in 2026, we still believe that we can formalise and encode everything before acting - despite the many failures in this mode.

I.e. Interview the experts. Build the process. Document the procedures. Then execute.

It's seductive by having the illusion of control.

But it assumes experts can articulate what they know and they aren't faced with conditions they haven't seen before.

But in my experience? The richest knowledge surfaces in the moment of decision.

Not weeks later in a workshop.

What if we flipped it? 🔄

Start with source material and goals. Let AI extract semantics directly. 

Capture tacit knowledge when SMEs override decisions, in the moment, with full context.

Let the system learn as it runs learn from it's decision traces!

Decision traces don't require predefined procedures to trace against. The trace is the execution. State, knowledge, reasoning, intent, captured in the moment. 

The 'why' surfaces through action, not up-front modelling.

SME & algorithm working together, capturing emergent behaviour. 

Bill Schmarzo champions Propensities that model actual behaviour algorithmically.

Mark Stouse proposes Causal models that understand what actually drives outcomes.

Adapting as reality moves - This is the shift 💡

I know it's frightening. But it's not pixie dust. It's happening and it's real.

From Legacy: Define then Execute. Massive cost and friction, upfront & change.

To Emergent: Capture algorithmically, execute then learn. Change is the operating model.

The hard question
In our excellent conversation, Henrik Göthberg asked: "Do we qualify out the ones not ready? Or can we bridge the gap?"

The companies that get this mode, exist. Revolut. Tesla. SpaceX. Starling.

They bake in innovation and change as a feature, not something bolted on top.

Note: If we had to define everything upfront, the start-up world wouldn't exist.

The CEO doesn't want a perfect model of today. They want a system that can handle tomorrow.

The flywheel turns. The knowledge lives, adapts, and delivers.

Nebulyx AI

AI - Dynamic Emergent Systems: The Flywheel Turns

The Knowledge Revolution Is Happening Now

Saturday reflections. 🤔

Something is shifting and I can feel it in the conversations.

The old world is creaking

Knowledge was something you captured upfront by interviewing experts, building ontologies, and hiring specialists for years.

Meanwhile, someone with 25 years of experience has six weeks left in the company. 

Someone shows up with half an hour to capture what they know. Sign off and gone. Maybe 40% captured if you’re lucky.

The tacit stuff never makes it because people can’t articulate what they do unconsciously. 

It only surfaces in the doing.

What organisations are actually asking for:

I’ve had several conversations recently with organisations facing this challenge. 

Offshoring programmes, retirement waves, restructuring. Knowledge walking out the door.

They don’t want another ontology project. 

They want AI that helps them understand what they actually do and how they do it. 

Deterministic, explainable, and traceable back to source.

They want to capture knowledge as work happens.

AI that accelerates pulling stuff out of their brain.

They want business people able to tune and tweak without waiting for specialists. 

They want knowledge that survives when people leave.

A new generation is answering the call:

What’s encouraging is that a new generation of AI knowledge solutions is emerging to meet this need.

We’re building this at Nebulyx AI with our Knowledge Flywheel and others are converging on similar approaches from different angles. 

Process scaffolds capturing rationale at every decision point. 

Context graphs as decision control planes. 

Deterministic agents grounded in domain truth.

Small models running locally, giving the same answer to the same question every time, with full provenance showing exactly where it came from.

This isn’t theoretical anymore. 

It’s real, working and it’s answering the call of organisations that need agile, flexible, and fast approaches to knowledge capture.

The shift

From design-time to runtime. 

From define-then-execute to execute-and-learn. 

From knowledge as library to knowledge as living executable system. 

From years of upfront work to agile capture as we go.

The knowledge revolution is happening now.

The flywheel turns. The knowledge lives.

Happy Saturday.

Nebulyx AI

No alternative text description for this image

AI Native Processes: Building Learning Systems: 🤔

I've been talking for a while on how we build processes in organisations and why the traditional approach keeps failing us.

The old separation
Process management said: document the flow first, then execute against it.

Knowledge management said: capture the knowledge first, then apply it.

Both assumed you could separate the knowing from the doing, defining everything upfront then running it.

But that separation was always artificial and why so many transformation programmes stall. 

Years of documentation with flowcharts that don't match reality on first contact. Knowledge bases nobody uses. 

By the time you've defined everything, the world has moved on.

AI Native processes don't start with complete documentation or perfect knowledge bases. They start with goals and domain knowledge, then learn through execution.

I've found 70-80% of processes being fundamentally the same across organisations. Procurement is procurement. Onboarding is onboarding. 

The variation is often just regional quirks & org chart contortions, not meaningful difference.

It's also amazing how many processes are part of the big LLM models training set. That's embedded knowledge we can draw on.

When we design processes with business people, we don't ask them to document everything first. 

We use AI to generate flows based on domain knowledge, industry patterns, and outcomes. 

Business people respond yes or no, they shape and refine for their specific context, and the process emerges through collaboration not committees.

What used to take months can happen in hours.

It's also interesting that Agents are not just execution engines, they are learning units.

Our agents operate within a bounded decision spaces, grounded in domain knowledge. 

As they execute, they captures every decision with full context: state, constraints, reasoning, outcome.

That's not just audit trail. That's the raw material for learning (this where Context Graphs come in).

New pattern emerges? The agent surfaces it. 

Edge case handled differently? Captured & available as precedent. 

SME overrides the agent's recommendation? The judgment is recorded in the moment, not lost in someone's head.

Traditional processes were statically defined by people long before implementation. 

Document once, execute forever, update occasionally when someone notices it's broken.

But agents learn from their own execution with the process getting smarter as it runs.

Process management & knowledge management were always artificially separated. Different teams, different tools, different methodologies.

AI Native collapses them back together.

The process is the knowledge in motion. 

The knowledge is the process made explicit. 

Agents capture both through execution.

Build fast with AI. Learn continuously through action. 

Every agent is an emergent system that compounds knowledge as it runs.

That's the shift , happy Sunday.

Nebulyx AI

AI Native Processes: Building Learning Systems

Context graphs: state, knowledge, decision - truth in the moment

After more excellent conversations about context graphs, I’ve been reflecting on why there’s so much “heated” debate. 🤔

One key aspect keeps surfacing.

The difference between point-in-time knowledge vs abstracting concepts, definitions, and flows ahead of time.

Traditional knowledge management: 📚
Abstract meaning out of context into universal definitions.

Define flows, ontologies, schemas upfront.
Maintain one “true” meaning of terms.

Problem: meaning drifts, flows break, maintenance never ends.

Solution: governance cycles, tracking drift, updating definitions. ⏳

Context graphs: 🎯
State, knowledge, and decision captured as one, in the moment.

Each decision trace is temporally bound.
No universal definition to drift.

Flows emerge through execution, not upfront design.

Example: customer discount approval 💰

March 2024: Customer gets 15% discount.

At that moment: spend history £50k, policy allowed up to 15% for that tier, sales VP approved, Q1 targets were behind. ✅

January 2026: Same customer, same request.

But now: spend dropped to £30k, tier changed, policy updated to max 10%, different approver required. ❌

Same customer. Same ask. Different truth in each moment.

Traditional KM asks “what’s our discount policy?” (universal, abstract).

Context graphs capture “what was true when this decision was made?” (contextual, temporal).

This is why Foundation Capital sees context graphs as organisational memory. 🧠

Memory isn’t abstract.

Memory is point in time.

What happened, when, why, with what information.

Traditional knowledge management tries to maintain universal truth.

Context graphs capture truth in the moment. 💡
That’s the paradigm shift.

Not just a new technology.

A different way of thinking about knowledge itself.

The Knowledge Flywheel doesn’t fight drift.

It captures evolution naturally through execution. 🔄

Nebulyx AI

taplio

The Battle

Jeff Frick

Jon Cooke

Followers

Posts

Engagements

Likes

Performance ScoreJeff Frick's LinkedIn Performance

Performance ScoreJon Cooke's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Jeff Frick

Jon Cooke

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);}Jeff Frick's LinkedIn Performance

Performance ScoreJon Cooke's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreJeff Frick's LinkedIn Performance