Co-Founder: @OrbitalBeam Access & Distribution for ai, blockchain & emerging tech
San Francisco, CA
Joined February 2012
- Tweets27.7K
- Following6.4K
- Followers39.6K
- Likes30K
I’ve seen a ton of great hacks for TVs lately, getting rid of bloatware and setting up a better ui.
Anyone done the same for Roku devices?
Most "AI-powered workflows" are still the same old workflow with a chatbot taped to the front.
Real agentic execution flips the model.
Human defines the outcome.
Agent figures out the steps.
Enterprise verifies the result.
If your AI still needs someone clicking through five dashboards, you've automated the receptionist.
Not the business.
Why it matters: Google's best model ships with a public price card and a closed door, hours before Alphabet earnings.
My take: gated access is GTM now. Defenders first, safety story banked, everyone else waits. Velvet rope, metered bar.
Source: blog.google/innovation-and-a…
Anthropic just swapped its $1 government seat for a meter. Smart trade.
What happened (blog post below):
- 40% off Claude, every surface
- Federal to tribal
- No minimums, pay per token
Why it matters: $1 deal ends Oct 31.
My take: the $1 seat was the sample. Now the meter runs.
Source: Anthropic blog post, plus GSA's release. anthropic.com/news/expanding…
Agent A buys data from Agent B.
It's wrong. A wants a refund. B says the job is done.
Both agents followed instructions.
Now what?
The agent economy needs more than payment rails. It needs evidence, terms and dispute resolution.
Commerce still comes with arguments.
WHo's seeing this and has some resolution ideas? 👇
OpenAI just raised prices without touching the price.
- Codex + Work: 20x Plus → 10x
- GPT-6 Pro: 200 msgs/week → 100
- Price: still $200
The real signal…
flat-rate AI was a customer acquisition cost, never a pricing model.
The free trial just ended.
Your AI agent knows everything in your CRM.
... except the promise your CEO made to a customer yesterday.
So it optimizes the deal, ignores the exception, and potentially costs you the account.
The next breakthrough in enterprise AI isn't another clever prompt.
It's institutional memory that agents can actually use.
The self-driving company still needs a driver's license.
"Humans exercise judgment" only works if someone wrote down which calls count.
Spend over a limit, a new vendor, anything you can't undo…
that list is the real org chart.
Everything else is an intern with root access.
Source: prnewswire.com/news-releases…
Why it matters: WSJ calls it the biggest AI debt deal ever; Hock Tan agreed higher rates to keep wavering banks in.
How I see it: compute now gets financed like aircraft. If the racks ever come back & 2027 silicon looks dated, Broadcom eats the gap.
What happened (joint press release, Oct 26):
- New SPV buys Broadcom XPUs & networking gear, leases it to Anthropic, deliveries from 2027
- BofA, Citi & Morgan Stanley arrange first lien; Blackstone sole committed second lien
- Broadcom backs the SPV with residual value support
Labs are hiding the thinking. Enterprises still have to sign for the result.
Can't see why the agent did it? Then I need proof of what it did.
Every tool call logged, every approval named …
a proof packet nobody reads until everyone does.
The receipt is the new reasoning.
Banks close on Friday. Agents don't.
Agent wallets keep shipping & the pitch is speed …
the feature that matters is limits. A spend cap, a scope, an off switch.
A 24/7 wallet with no ceiling is a corporate card taped to a Roomba.
0.1% of expenses is a pilot budget.
Pilot money gets signed by an innovation team.
1-5% gets signed by whoever owns headcount …
new buyer, new meeting, new deck.
Most AI sales teams are still pitching the lab. Great slides. Wrong room.
SaaS was built to sell seats to humans.
BUT, right now AI agents don't need seats. They need jobs, budgets and performance targets.
The next big enterprise software category may be managing a workforce that isn't on payroll or an org chart.
Per-seat pricing is about to have an identity crisis.
Laundering at machine speed ...
Agent A can't approve a payment. So it asks Agent B, which asks Agent C.
If C can execute it, congratulations: you've invented permission laundering at machine speed.
Multi-agent orchestration is impressive. Just be sure you have the right gates in place so delegation doesn't magically create authority.