Estimated Reading Time: 9 minutes
Contents
- Preamble
- On AI & The Overton Window of Weirdness
- On Strategy: Rules of a tether
- On Tethering to an exponential
- On Execution
- On Attention Systems
- On Intelligence Engines
- On Agent Chains
- On Competitiveness
- On Accountability
- On Meaning
"The pace of change is so fast that soon enough visionary will become a lifestyle, rather than a breakthrough thing." - Salim Ismail
Preamble:
It's been a long time since journaling; Two years, and everything has accelerated. In that time I've been mentored by world-class operators in AI, Data, Cybersecurity, Strategy and Transformation. They rewired how I think about all of knowledge work & professional services.
Along the way I've worked on strategy to keep up with AI and it worked.
I became the AI SME at a major Professional Services Firm and a Finalist in three enterprise categories for the 2025 Australian AI Awards (Leader OTY, Consultant OTY and Rising star OTY, alongside impressive operators).
A key learning that matters for this piece: Strategic impact scales with massive action. Define Win Conditions, build an Attention System, and tether yourself to the complexity curve long enough to see the non-obvious edges. That’s how you convert an exponential into an advantage.
Soon I'll be stepping into an AI Solution Architect role with people I'm excited to build with and learn from.
Consider this post a trade: I'll share my proven strategy to meet you - the serious operators pushing this once-only moment in a humanist direction at the beginning of the Age of Intelligence.
AI and the Overton Window of Weirdness
In 2018 I watched OpenAI 5 play some of the worlds best Dota 2 players real-time. The bots showed mechanical skill, psychological pressure and even used a novel exploit mid-match. As a former Australian pro in real-time-tactics games, I felt the floor move, the writing was on the wall "A country of geniuses in a data centre" had been proven feasible.
"A country of geniuses in a data centre." - Dario Amodei
Not many knew it yet, but this small team of ambitious researchers had just blown open the Overton Window of Weirdness. Since GPT 3 I've treated AI as a domain where reality reorganises faster than our priors update.
"The Overton window of weirdness, is the range of things acceptable to spend one’s time doing, when it is narrow we are optimisers, and when it's wide a societal random walk occurs leading to discovery."- Matt Webb
Highlight of OpenAI Five - inventing an exploit to win,
Listen to the crowd roar to hear the magnitude:
On Strategy: rules of a tether
Fast Definitions
- Win Conditions: self-evident or measurable states that, if achieved, mean "we won".
- Axioms: explicit strategic bets about the interplay between how the world behaves and your win conditions.
- Plan: next actions contingent on the axioms of strategy.
- Tether: a designed Attention System that keeps you attached to an accelerating Complexity Curve long enough to find the non-obvious edges of any problem space.
Hard rules
- Write out your win conditions. If they aren't defined, it isn't strategy.
- Input mode beats willpower. Cognitive load shifts, pick appropriate mediums for how you feel on the day.
- Build an attention system. Compress entropy into signal with a repeatable stack. Take your trusted sources and work their content for digestibility.
- Publish artifacts regularly. Build tools, slide-decks, diagrams, write articles, attend events. Increase your luck surface area when solutioning.
- Trade IP for network. Collaboration compounds faster than hoarding.
Tethering to an exponential
Axiom A - Frontier problem convergence:
As experience in a field compounds, the problem space for experts typically narrows. Teams working on the frontier tackle near identical problems on similar timelines. Most options to solve a problem are not yet possible, or too expensive; a small number of options remain. Advantage accrues to those who stay on-task in the problem space the longest.
Axiom B - Complexity scales with exponentials:
AI progress is now exponential; capability is scaling across every industry. The complexity curve is rising in lockstep with each improvement, and individuals are outpaced in every industry unless they build Attention Systems that denoise the release landscape and stay linked to macroeconomic signals.
In this high-complexity state, non-technical executives are overwhelmed, and their strategic axioms drift out of alignment with future win conditions. The remedy is explicit: tune your Attention Systems to remain tethered to the curve as long as possible. Even if you’re eventually bucked off, you will see more and sooner than peers who let go early.
On Execution
"Advantage accrues to those who stay on-task in the problem space the longest."
Assumptions:
1. Move quickly to accumulate deep AI experience in the target area.
2. Design highly leveraged Attention Systems that keep us with the problem for as many hours per week as possible.
→ Result: We stay tethered to the exponential and pull our thinking far ahead of the status quo.
Core question:
How do we manage cognitive load while increasing our weekly hours on the real problem space (vs passive consumption)?
Peers average ~10 hrs/week; aim for ~30 hrs/week by restructuring your input modality dynamically.
Plan (don't fight your attention or schedule):
Some days you won’t want another AI report. Others, a documentary, podcast, or YouTube breakdown will land better. Sometimes the move is hands-on: build an AI tool in Cursor or spin up images in Midjourney. Switch the channel, not the mission. Change how info enters, is processed, and is expressed to match your energy and schedule so you can stay in the problem space longer and actually master it.
Input modalities: Video, audio, text, images/diagrams.
Adaptive processing: Turn documents into podcasts, podcasts into mind maps, YouTube explainers into structured docs and convert any of the above into a live quiz with Voice AI. Use custom tools in Cursor or products like Google’s NotebookLM, OpenAI’s ChatGPT, or Copilot’s Voice Mode.
Output modalities: Publish small artifacts, notes, posts, quick diagrams, or even a structured chat with a colleague - to lock learning and create feedback loops.
Caveat (expect the buck): The curve will throw you eventually; knowledge will be superseded before your priors fully update. Your job is to extend time-on-curve so when it happens, you’ve seen more, earlier and are better positioned to anticipate what’s next.
AI Capability map (modality alignment)
Use these broad categories to define intake/processing/output modes needed to reduce cognitive load and acquire durable advantage.
· Generate: Create new artifacts from existing content.
· Transform - Convert content between formats or data types.
· Distill - Summarise, Key quotes, Concepts etc to increase digestibility.
· Find - Pinpoint critical information hidden in complex data.
· Reason - Analyse, Draw conclusions, chart decisions.
· Do - Automate action, execute tasks, Initiate Attention Systems.
· Represent - recreate characteristics or attributes of the person in speech, video, image, text.
On Attention Systems
Plain definition: An Attention System continuously pulls the right signals closer, at the right fidelity, with the least friction - so you can stay tethered to the complexity curve longer.
An Attention System is a small pipeline with five parts:
- Sources – frontier research, release notes, market news, internal docs, dashboards.
- Collectors – scheduled searches/feeds (web search api's, RAG over internal wikis, cloud docs, email digests etc).
- Processors – LLM transforms to Distill (summaries), Transform (format shifts), Find (targeted pull), Reason (quick takes/implications).
- Memory – lightweight store of highlights, decisions, and “watch items” (tags + timestamps).
- Surfaces – a single daily/weekly brief, plus live boards for “What changed?”, “Why it matters?”, “What do we do?”
Workflow example:
Say “Collate latest discussions about AI Security Research across all frontier labs such as Anthropic, Google, OpenAI, xAI”
The system runs queries across web API's, scrapes release databases, pulls internal deep research notebooks, distills to a 1-pager with links and a “so-what” section, and opens tasks tagged Next Bets. You skim, decide, move.
On Intelligence Engines
Plain definition: An Intelligence Engine is your multi-modal thinking workspace that supercharges critical thinking, ideation, aligned to core objectives. A “team brain” that builds noisy off-the-cuff thoughts into blue ocean strategy or detailed risk-adjusted next-steps.
The future of critical thinking is a team effort. As foundational AI models improve and humanity demands more from itself, knowledge workers will step up an abstraction layer - raising expectations of quality.
I’ve already built much of the following, albeit haphazardly, across emerging workflows. Unification into a single engine becomes feasible as frontier gaps close and the operating landscape stabilises.
(Crude historical analogy: just as Assembly → BASIC → C → Python → Rust/Go each pushed programming up an abstraction layer, Intelligence Engines push knowledge work up a layer. You still think - just with better leverage.)
An intelligence engine might be made of four layers:
- Intake – voice/text/file drop (meetings, briefs, specs, data excerpts).
- Retrieval & Contexting – pull internal knowledge, metrics, prior decisions; attach provenance.
- Reasoning Models & Tool-use – orchestrated prompts/agents to Distill, Compare, Scenario plan, Critique, Translate modality (e.g., doc → diagram → mind-map).
- Artifacts & Memory – outputs as decision memos, PRDs, diagrams, mind-maps, plus an indexed reasoning trail.
Humans still need to think in the Age of Intelligence. Companies may soon view talent as unique contributors to the Intelligence Engines they use - hiring not only for role fit but for the measurable lift a person brings to the organisation’s “team brain.” As Satya Nadella has suggested, employees may effectively be hired with their agents. In that world, experience across Intelligence Engines - and the ability to improve them -raises the value of prior work, and can broaden pathways for diverse talent whose strengths become legible in shared reasoning systems.
Workflow example:
Speak into the engine or prompt it and it ideates alongside you, sustaining longer flow states. It bounces ideas back, reframes problems through different critical-thinking tools, and translates your rough notes into structured options (with sources), raising the resolution of emerging ideas.
On Agent Chains
Plain definition: Agent Chains are outcome-oriented automations that coordinate tools and approvals - under explicit GRC guardrails - to complete real, risk-adjusted workflows, then hand results back to humans for report or review.
What:
Bring approvals and actions to the user’s fingertips. Surface only what requires attention. Execute the rest safely in the background.
Why:
Handle the work people must do (comms distribution, distillation, data in/out of databases) without swivel-chair effort.
Orchestrate end-to-end actions across other agent systems.
Produce clear outcomes that a human can approve, reject, or amend.
How:
Integrate systems so agents can coordinate tasks across the stack: Microsoft Graph, SharePoint, Xero, CRMs/ERPs, data warehouses, OpenAI Enterprise, Claude, Google AI Studio, Perplexity Labs, and more - each with least-privilege scopes and auditable trails.
Modern work adjustments:
Users receive or initiate actions that pass cleanly between human and agent teammates for further review.
Attention Systems feed fresher signals in; Intelligence Engines work the data into blue ocean strategy alongside you; Agent Chains carry approved actions out. Same tether - three layers of leverage.
On competitiveness: a brief note
In any strategic game, advantage boils down to decision speed, decision quality (impact on Win Conditions), and how those two variables constrain a competitor’s options. All of it sits under the clock: time governs speed, and it governs quality because real quality in ambiguous domains comes from iterations and reps that sharpen your read on the Win Conditions.
The tethering rules raise your odds of winning in any industry.
Remember: in private enterprise you compete with firms and with every individual seeking the same roles. No sector is insulated. Even trades touted as “future-proof” will absorb disruption as automation reshapes the job mix; displaced workers will retrain into whatever remains economically viable.
Apply the tethering rules to your field now. Build the advantage for your team and for those who depend on you.
On Accountability
How do you know you’re ahead of the curve and not falling off?
Anchor yourself to reality with a community of visionaries, shared artefacts, and discussion with outside voices, trusted technical opinions, macroeconomic signals.
Community checks: publish small artefacts, invite critique, compare notes with credible peers.
Aggregated expertise: track consensus and dissent across trusted evaluators; log what changed and why it matters.
Open trading of ideas: treat selective IP sharing as a hedge collaboration compounds faster than hoarding.
Accountability is the tether’s safety line: it keeps you honest when the curve gets steep.
On Meaning
"Each person is responsible for all people and for all things." - Brothers Karamazov by Dostoevsky.
"It may be possible for each to think too much of his own potential glory hereafter; it is hardly possible for him to think too often or too deeply about that of his neighbour. The load, or weight, or burden of my neighbour's glory should be laid daily on my back, a load so heavy that only humility can carry it, and the backs of the proud will be broken." - The weight of Glory by CS Lewis
If you do not take it upon yourself to build AI solutions for your fellow neighbour that encourage human flourishing who will? Even when incentives pull the other way, you should try. Progress is hard and markets often reward short-term self-interest; act anyway. The work is to align capability with consequence: design systems that make it a competitive advantage for companies that do the right thing, and harder to be anti-humanist, at scale.
"Fight hard to build the best future for all of us. We get to invent these first-conditions of our economy - living in the Age of Intelligence - just one fragile time. Ethical solutioning for executives, politicians and technical employees is critical, before third-degree path dependence and institutional lock-in causes a permanent underclass to be entrenched. The need for a big beautiful vision is paramount to our success. We can do better than this. Please fight for it." - Samson Blackburn
If sending a connection request, only those with messages that introduce your intentions in this space will be accepted: https://www.linkedin.com/in/samsonblackburn
Top comments (0)