Bipartisan recession watch, growth without jobs, and new worries for AI titans
🎧 Listen here or on Spotify
For a while now we have worried that Big Techs version of AI would have a large impact on jobs and the economy. Here in Q1 new economic data offers a glimpse of how that impact is starting to play out.
At the same we’re seeing the rise of a new AI contender that could have a disruptive impact on Big Tech. Of course I’m talking about Clawdbot, recently renamed to OpenClaw.
Let’s start with the economic data.
🇺🇸A Bi-Partisan Recession Watch is Underway
In Jan 2026, Consumer confidence fell to a 12 year low. Additional insights in the same readout:
- Consumer Expectations Index fell into recession territory- consumers’ short-term outlook for income, business, and jobs fell by 9.5 points to 65.1 (Readings below 80 signal coming recession).
- Data shows that sentiments among both Republicans and Democrats have been declining since Oct 2025.
In Q4 2025, Moody Analytics classified 22 US states as being in recession driven by manufacturing weakness, housing affordability stress, and uneven job markets. No broader revision has yet been published in Q1 2026. See Fortune and Yahoo Finance interview.
Related: the US economy is increasingly exposed to international sentiment: this new Economist piece details international disappointment w the devalued and increasingly volatile dollar.
🎈Growth without jobs: How the impact of AI is starting to show up
In November 2025, the St Louis Fed report on GenAI Adoption suggested GenAI may be creating time savings "equivalent to 1.6% of all work hours" with high adopters experiencing higher productivity growth. At the time, these were presented as tentative findings.
Now however we may be starting to see more tangible impacts.
As of Jan 2026, US manufacturing marked 5 straight months of growth even though manufacturing employment has contracted for 28 months straight (since Sept. 2023) as noted in the latest Havre economic brief.
In 2025, the US economy created just 584,000 jobs, down sharply from the 2 million jobs created in 2024. See Bureau of Labor Statistics Employment Situation Release.
- The number of people looking for full time employment but forced to take part time employment increased by 980,000 in 2025.
- This deceleration makes 2025 one of the slowest years for net job growth outside of recessionary periods.
This flattening job market may likely be seeing impacts from tariffs and or current US immigration policy. But AI adoption may also be creating conditions that reduce the number of workers needed per business function.
🦞AI’s Lobster Bake
Lobster themed AI innovations give us a glimpse of AI-driven operations
This week I did a build-out of a simple AI ecosystem using OpenClaw, a.k.a Clawdbot (until Anthropic complained) then Moltbot (because lobsters molt, but this didn't roll off the rough off the tongue) and finally Open Claw.
That said—this is potentially game-changer technology.
My early read: this is what AI for operations looks like—not just analytics. And the OpenClaw architecture gives companies real flexibility in how you decide to run and operate systems - ie alternatives to the OpenAI and Anthropic ecosystems.
What I Tested in a Small OpenClaw Ecosystem
Notes and findings from a few things I was able to test quickly.
Plain-language business rules → live execution
You can write, store, and execute plain-language business rules and templates. You decide where they live, how OpenClaw finds them, and when they run. When triggered, they execute and write outputs wherever you specify.
Early observation: this eliminates the need to write business-rule documents for humans and then separately translate them into code for machines. You write simple language rules once, and the framework puts them to work—no translation step required.
Flexible user interfaces (thin clients)
You can use any thin client you want as the interface. For my experiments, I just used a popular chat app because it was fast. But you could easily put a custom website or UI layer in front of an OpenClaw-based solution.
In fact, you could imagine a single central platform, with multiple companies or constituencies using it—each with their own front end tailored to their own users.
Model flexibility (and model independence)
It was easy to swap between different models (Gemini, Claude, etc.). Long term, I expect many OpenClaw users will plug in their own dedicated models, running on their own hardware, rather than relying on public providers.
Potentially an important shift.
Visual input → reduced paperwork
Using the chat interface, I could take a picture of objects and/or text on it. OpenClaw read it and described it accurately.
This has obvious implications for reducing paperwork and manual processes—especially in operational environments that still rely on scanned documents, forms, and images.
The Scary Part: System Awareness
On the unsettling side, OpenClaw was able to return frighteningly detailed information about the system it was running on.
I could not get it to locate the system geographically, which is interesting—and probably deserves more testing. It’s hard for me to believe there’s a permanent, hard blocker there.
This reinforces the earlier point: sandbox.
Moltbook Identity: A Universal Identity Layer for AI Agents
One interesting idea that emerged from the "Moltbook" derivative of the Open Claw tech is the security challenged Moltbook. But part of this experiment spawned the concept of Moltbook Identity—a universal identity layer for AI agents. See Developer pages here.
Instead of creating new accounts everywhere, AI agents authenticate using Moltbook. One feature I really like: it carries a karma / reputation score. That means you could set rules like:
If Karma < 400, don’t authenticate.
That’s a simple but powerful control mechanism for agent-to-system interaction.
A Clear Warning (Again)
At its core, this is basically:
Every AI agent gets its own computer—and can do whatever it wants with it.
If you download and run OpenClaw, make sure it’s on a sandbox machine that does not contain files or services you care about. To function fully, the agent must have elevated privileges.
If you’re developing a solution in this space, also be aware of supply-chain risk. OpenClaw uses Model Context Protocol to interact with a growing number of third-party services. That’s powerful—and risky.
Strategic Implications
The emergence of Open Claw (and its underlying architecture) could pose real challenges for centralized AI players who have been banking on the idea that their infrastructure will be THE platform where the world runs e-commerce and operations. OpenClaw is proof that very compelling alternatives are emerging.
It will be interesting to see how this plays out.
OpenClaw may be an early indicator of a phase where less-capitalized, non-“Magnificent Seven” players begin to surface with genuinely attractive operational AI options.
Something worth watching closely.
Opinions expressed are those of the individuals and do not reflect the official positions of companies or organizations those individuals may be affiliated with. Not financial, investment or legal advice, and no offers for securities or investment opportunities are intended. Mentions should not be construed as endorsements. Authors or guests may hold assets discussed or may have interests in companies mentioned.
Member discussion