SORRYWECAN × Research Lab · Mindshift 2026
Imagination and Clarity
are the only bottleneck.
A high-level orientation on what's shifting, how to think about it, and what's actually possible right now — from someone building in it every day.
Agent Orchestration
Future of Work
New Thinking
Adaptability
01
01 · The Shift
This isn't a new tool.
It's new physics.
AI doesn't automate tasks. It reorganises who can create what. The disruption isn't happening in one industry — it's happening in the architecture of work itself.
39%
of current skills will be transformed
by 2030 · WEF
by 2030 · WEF
75%
of knowledge workers already
use AI · Microsoft 2024
use AI · Microsoft 2024
300M
jobs projected impacted
by 2030 · Goldman Sachs
by 2030 · Goldman Sachs
9×
surge in AI-skill vacancies
since 2022 · World Bank
since 2022 · World Bank
These numbers don't demand panic. They demand orientation.
01 · The Gap
AI can do far more
than it's being used for.
Anthropic research: the gap between theoretical AI coverage (blue) and observed usage (red) is enormous.
Knowledge work — management, business, finance, legal, computer — is theoretically near-fully automatable. Almost none of it is being automated yet.
The window is open.
Those who move into the gap now build an advantage that compounds.
Source: Anthropic — Labor Market Impacts of AI, 2025
01 · The Shift
This isn't a new tool.
It's new physics.
It's new physics.
AI doesn't automate tasks. It reorganises who can create what.
Research Lab
02 · Systems Thinking
AI is not one topic.
It's four systems — looping.
01 · Labor
Work is restructuring
Not disappearing — redistributing. Who does what, how fast, at what cost. The map of competence is being redrawn.
02 · Education
Credentials are losing ground
What's worth learning is shifting faster than curricula can update. Agency and adaptability beat memorised knowledge.
03 · Capital
Compute is the new land
Whoever owns the inference stack owns the leverage. Top 1% of compute owners will pull further ahead — everyone else rents their tools.
04 · Creation
One person = entire studio
Film, product, code, research, brand — a single focused person with agents can now do what required a 10-person team two years ago.
These four reinforce each other. Understanding the loop is the edge.
AI
Loop
Loop
Labor
Who does what
Capital
Who owns compute
Creation
One person = studio
Education
What's worth learning
02 · The Loop
Four systems.
One loop.
Each domain reinforces the others. Labor shifts change what education is worth. Capital concentration shapes who can create. Creation redefines what work means. The loop accelerates.
Understanding the loop is the edge.
03 · New Operating System
The shift isn't in the tools.
It's in the thinking.
Old Way
What tool do I need to learn?
Can I execute this myself?
Is this technically possible?
I need a team for this.
Creative work needs resources.
→
New Way
What do I want to create?
Can I describe this clearly enough?
Is this worth imagining?
I am the team. Agents execute.
Creative work needs a clear idea.
Research Lab
04 · The Skills That Last
What AI can't replace.
What to build now.
⚡
Adaptability
The ability to update your model of reality faster than the world changes it. The most valuable skill of the next decade.
🧭
Agency
Initiating without permission. Deciding without certainty. Moving before the full picture is clear.
🔍
Curiosity
The engine that drives better questions. AI amplifies curiosity — it cannot replace it.
🪞
Self-Knowledge
Knowing what you want to build, what you're optimising for, and what you're willing to delegate.
⚖️
Judgment
Deciding what matters. AI produces. Judgment selects. This is the skill that cannot be prompted.
🌐
Systems Thinking
Seeing loops, not events. First, second, and third-order consequences. The mental model that doesn't expire.
AI builds fluency fast. These six are the foundation it builds on. They take time — and that time starts now.
05 · The Framework
You are the director.
Agents are the execution.
Imagine
Define what you want to exist. Not vaguely — with enough clarity to describe it to someone who has never seen it.
→
Decompose
Break the vision into tasks. Which parts need reasoning? Which need speed? Which need creativity?
→
Route
Match each task to the right model. Claude for structure. Gemini for language. DeepSeek for depth. Haiku for speed.
→
Verify
Don't accept the first output. Push back. Ask the uncomfortable follow-up. Quality lives in iteration.
→
Ship
Small and done beats ambitious and perpetual. The first version is not the last version.
The hardest part isn't building. It's imagining clearly enough to ask.
05 · What Stays Human
You own the beginning and the end.
You (Input)
🧠
Imagination
What to build, why it matters
🎯
Clarity
Precise enough to ask for it
🎬
Direction
Set the goal, the tone, the constraint
→
Agent Execution
✉️
Email outreach
📊
Research & synthesis
💻
Code & build
📝
Drafts & copy
📡
Monitor & scan
🔄
Pipelines & workflows
→
You (Output)
⚖️
Judgment
Is this right? Does it land?
✅
Decision
Ship it, change it, kill it
🔁
Iterate
Feed back, improve, go again
The machine executes. You stay in the loop — at the parts that matter.
05 · You Are the Director
You set the direction.
Agents execute.
Agents execute.
The intelligence is in the orchestration, not the output.
Research Lab
06 · Agent Orchestration
Orchestration is the new literacy.
Decompose the problem
Break any task into its components. Identify what requires reasoning vs. what requires speed vs. what requires creativity.
Route to the right model
Each model has a different relationship with language, logic, and depth. Knowing which to use when is the skill.
Iterate in conversation
One prompt is not a workflow. The best outputs come from a series of sharpening exchanges — like a real collaboration.
Build memory and context
Agents that know your projects, preferences, and rules compound over time. Context is the moat.
Live Routing Table — How I Route Tasks
Frontend · UI · animations · React→Gemini 3.1 Pro
Blog posts · copy · Slovak translation→Claude Sonnet
Deep research · papers · analysis→DeepSeek R1
Pipelines · cron · automation · data→Claude Haiku
Agent orchestration · architecture→Claude Opus
07 · What Runs While I Sleep
Every day, a pipeline
runs automatically.
08:00 · Daily
Morning Brief
10+ sources synthesised. Intelligence brief published to site. Radar updated. Delivered to Telegram.
07:30 · Weekdays
Finance Intel
Live portfolio from Alpaca. X signals from 30 accounts. Finance brief — DM before the market opens.
11:00 + 17:00
Trending Feed
X scanned for viral AI / tech posts. What's moving in real time — surfaced before it becomes a headline.
18:00 · Daily
Daily Reads
Best long-form from 10 RSS feeds. Philosophy, capital, AI, culture. Scored by signal quality, not recency.
Sunday 19:00
Weekly Synthesis
Full week of signals synthesised. 3 patterns with 1st / 2nd / 3rd order consequences. EN + SK.
R&D · 10:00 + 18:00
R&D Council
5 AI personas debate my projects twice daily. Scout, Architect, Engineer, Growth, Creative. Auto-enhance queued.
Research Lab
08 · The Real Edge
Clarity is the bottleneck.
Everything else can be built.
1
Curiosity→
2
Right Question→
3
Right Model→
4
Deep Answer→
5
Build→
6
ShipAll the data, the code, the copy, the automation — it can all be built. The only question is whether you can picture it clearly enough to ask.
Prompt engineering is dead. Orchestration is the new literacy. Imagination — combined with clarity — is the only competitive moat that can't be automated away.
08 · The Real Edge
Clarity is the bottleneck.
Everything else can be built.
Everything else can be built.
All the data, the code, the automation — it can all be built. The only question is whether you can picture it clearly enough to ask.
SORRYWECAN × Research Lab · Mindshift · March 31, 2026
What would you build
with 10× more
execution capacity?
Not a better tool. Not a faster laptop. An agent that knows you, learns from you, and runs while you sleep. That's the question worth sitting with.
Research Lab
research-hub.xyz
Inšpirované: Citrini Research · Anthropic · WEF
10