Strategic intelligence for the AI era.
Nashville, TN
Joined May 2025
- Tweets24
- Following10
- Followers9
- Likes5
Pinned Tweet
The future isn't something you predict. It's something you participate in shaping. And the people doing the shaping deserve better tools than a Bloomberg terminal and a gut feeling. That's what we're building at Creative Foresight. More soon.
To run an effective team of AI agents your intent has to survive tool boundaries. You express your intent - and it must use many APIs, CLIs, and MCPs which all have a different 'shape'. The agent re-forms and conforms its actions to each tool, but does your intent survive? This is key for enabling agents to run for long periods of time unsupervised.
Cohort is your agent team that keeps your intent alive over the long term. Set up a Goal and see your agent team check in on progress and make multiple attempts to achieve it, from multiple angles, over long time spans.
Curious? Try it for free here: cohort.bot
AI can automate entry-level work faster than firms can redesign how people become senior enough to be able to steer AI system. Worthy of a new "Law" of organizational science?
So many people are new to coding, now that AI makes it approachable, so there are fundamental concepts that most coders learn early in their schools/careers that people now must learn at a much later point. Here's one: the Principle of Least Priviledge. Every unused interface, broad permission and open channel preserves optionality for the operator and opportunity for an attacker. Good security begins by deleting capabilities and permissions the mission does not require.
An agent skill description is executable routing logic. If it only explains when to activate, it will steal work from better tools. Define the cases where it should stay dormant as carefully as the cases where it should run. Restraint is part of capability.
Creative Foresight retweeted
The dollar’s "exorbitant privilege" of being the leading reserve currency comes with a downside. Consumers get cheap imports, but this imposes relentless pressure on American factories and exporters. About this, JD Vance is absolutely correct. But why does it have to be this way? 👇️
Every meaningful decision contains a forecast. Taking a job, starting a company and choosing where to live all reveal a private theory of the future.
Strategic foresight begins when we make those hidden forecasts explicit enough to test. And here's the thing - this is a skill that you can develop.
Follow this account to learn new actionable techniques for developing foresight. Life is good when you aren't caught off guard by unexpected new developments in the world!
Want to know where a market is stuck? Track its friction: abandoned checkouts, failed payments, compliance-killed deals, settlement delays and bid-ask spreads. Completed transactions show demand. Near-transactions show where value is being blocked.
Payment friction has more explanatory power for understanding markets than many realize.
Multi-agent work needs an enforceable task graph.
Define the owner, blocker, expected artifact and approval gate for every task.
Then prevent downstream agents from starting early.
Cohort treats dependencies as operating rules the team can execute.
Ship faster than you ever thought possible with Cohort – your ready-to-run agent team. Safe, sandboxed, reliable, and damn good at burning through your To Do list. cohort.bot
Generalists spot convergence. Specialists know whether it can ship. These are distinct but complementary – and what generalists bring to the table is often misunderstood.
A strong generalist researcher will spend 80% of her time inside two or three domains and 20% crossing history, culture and regulation. This is the best way to find high-order invariants that correlate to stable properties of reality. And understanding reality is how to make strong decisions in complex data-rich environments.
Make better decisions by learning how to generalize across domains. Get in touch with us to learn how your organization could benefit. creativeforesight.io
Agents need money! And they have it now – with x402 for stablecoins and L402 for bitcoin, agents can make their own economic decisions. But agent wallets need *policies* too. Spend limits per task, approved vendors, rules for reversible purchases, receipt requirements and escalation thresholds. Autonomy becomes useful when financial authority is explicit.
We are building the best agent wallet solution in Cohort. Check it out! cohort.bot
At CF we noticed “enabling ratios” precede breakthrough headlines.
For example, in emerging technologies, we monitor:
- intelligence per watt
- energy cost
- data moved per watt
- power or thrust per kilogram
etc.
When several curves cross viability thresholds together, new industries appear quickly.
This is like saying markets are downstream from technology. Sounds obvious when you say it but most investors trade on vibes instead.
Technology teams are trained to remove friction. That instinct becomes dangerous when friction is doing useful work!
A confirmation delay can prevent error. An approval gate can stop abuse. The strategic question is which resistance protects the system, vs. which is truly undesirable and just a drag.
Strategic friction - new concept that companies, especially aggressively AI-first ones, are wrestling with.
The strongest AI model on your team should behave like a senior operator!
Give it the goal, constraints and review criteria. Let cheaper agents research, code and test.
Bring the strong model back for decomposition, judgment and final review. Allocation compounds capability!
AI makes interfaces cheap. It's wild that making a dashboard is about as easy as sending an email nowadays. The new bottleneck is everything behind them: auth, data, files, realtime events and model access.
Teams that package those primitives into a boring, zero-config layer will ship more real experiments than teams generating prettier demos.
That's what Cohort is all about. Coding, marketing, and more Get early access! cohort.bot
Agents can already own tasks and hand work to one another.
The hard part is operational trust. Of course you need explicit assignments, enforced dependencies, and audit trails. But you also need a clear identity. Not one God-agent that does everything. A team with defined roles. At Creative Foresight we believe specialization of labor should and will carry over to the agentic era.
Specialization of labor is the foundation of a serious human-agent team.