🏀 Agent arena | Axios AI+ Summary

1. Key Themes

Theme: The agent "last mile" problem: capability is outpacing usability and trust

Agents can execute complex work but stumble on real-world transactions

Both Dots and Muse handled research-heavy, multistep tasks well, but actual purchasing exposed friction.
Quote: "Both agents were surprisingly good at handling complicated, multistep assignments, such as looking up flights, finding a sign-up for a ticket presale and flagging a personal conflict."
Quote: "Muse attempted to complete the purchase but got stuck at Target's sign-in page. Dots instead directed me to the Target app, where my account and payment information were already saved."

Less automation can mean a better experience

The less "autonomous" approach won because it leveraged existing authenticated state.
Quote: "Dots' less automated approach turned out to be faster, even though both services support purchases with user approval."

Authentication, permissions and human-in-the-loop steps are the structural bottleneck

The Hark episode shows the issue is industry-wide, not product-specific.
Quote: "AI agents can navigate websites, fill out forms and complete purchases, but doing so often requires authentication, permissions and human intervention."
Quote: "Those extra steps can make an otherwise simple transaction more complicated and, at times, easier for a user to complete on their own."

Trust is the binding constraint on autonomy, not just capability

Quote: "Both agents could have done more if I'd been willing to give them greater access to my accounts and wallet. I'm not yet comfortable authorizing purchases of hundreds or thousands of dollars, or at a point of trusting them enough for consequential decisions."
Quote: "For me, the sweet spot for these agents is doing the research and legwork while leaving me to make the final decisions."

Theme: Power is the binding physical constraint on the AI buildout

Data center power demand is set to surge against slow infrastructure

Quote: "Goldman expects global data center power demand to surge 170% between 2025 and 2030, with more than 60% of that growth coming from the U.S."
Quote: "Power is probably the forever bottleneck," said Jason Tofsky, Goldman's global head of digital infrastructure investment banking.

The scale of capital at risk is enormous

Quote: "The bank estimates roughly $7.6 trillion in capital spending between 2026 and 2031 across computing, data centers and power infrastructure."
Quote: "Goldman points to power as the biggest risk to trillions in spending."

Interconnection queues are the gating mechanism

Quote: "The U.S. has roughly 2,600 gigawatts of proposed power generation and storage projects awaiting grid connections, more than twice the grid's existing installed capacity, though many projects may never materialize."
Quote: "The median time from an application for a grid connection to commercial operation is roughly five years."
Quote: "Silicon Valley moves fast and breaks things, but the U.S. electrical grid moves slowly and requires permits."

Theme: Personal AI agents are becoming a competitive consumer category with distinct pricing strategies

Meta and OpenAI are taking different go-to-market approaches

Quote: "Muse is free for most uses, with optional $20 and $100 monthly subscriptions. Dots requires a ChatGPT Pro subscription, starting at $100 monthly, or access through a business or enterprise plan."
Quote: "AI companies are racing to create the digital assistant we turn to for everything from shopping to travel planning."

Infrastructure players are positioning around agentic commerce

Quote: "Agents, merchants, issuers and platforms need a common way to exchange information." (Sponsor content, Synchrony)
Quote: "Synchrony is helping shape open standards that can carry product, financing, rewards, permissions and transaction information across AI shopping experiences."

2. Contrarian Perspectives

More agent autonomy doesn't necessarily mean a better user experience

Consensus assumes fully autonomous agents are the goal. The evidence suggests removing human steps can backfire when it triggers sign-in walls and failures.
Quote: "Dots' less automated approach turned out to be faster, even though both services support purchases with user approval."
Quote: "AI agents are getting better at doing things for us. But that doesn't necessarily mean they're making our lives easier."

The bar for agents isn't capability; it's beating the human doing it themselves

Quote: "AI agents promise to take everyday tasks off our plates, but they still have to clear a basic hurdle: making a task easier than doing it ourselves."
Quote: "They also need to pass what we'll call 'The Human Benchmark': that is, can they do something as fast, or as well, as a human? The answer is ... not quite yet, but the capabilities are still impressive."

AI-generated writing may be eroding established operating rituals (and spawning agent-on-agent communication)

Quote: "A viral Substack post points out that Amazon's famed six-page memo is no longer as useful because it's been turned into AI slop."
Quote: "One engineer began tagging AI-generated memos with 'AI;DR' to indicate he wouldn't be reading them. His agent could just read and summarize those instead."

3. Companies Identified

OpenAI (Dots)

Description: AI lab; maker of the Dots personal AI agent.
Why mentioned: Tested head-to-head; its agent produced the more visually engaging deliverable and the faster purchase path, but at a higher price floor.
Quote: "Dots' 38-page vintage program was more visually engaging, with player photos, graphics and extensive statistics."
Quote: "Dots requires a ChatGPT Pro subscription, starting at $100 monthly, or access through a business or enterprise plan."

Meta (Muse)

Description: Tech giant; maker of the Muse personal AI agent.
Why mentioned: Tested head-to-head; free for most uses, showed proactive self-correction but failed at checkout.
Quote: "Muse is free for most uses, with optional $20 and $100 monthly subscriptions."
Quote: "Both programs contained questionable statistics. Muse later sent an unsolicited update correcting several factual errors."

Hark

Description: AI startup building a consumer agent (Hark Pro).
Why mentioned: Case study in agent scaling failure during a promotional stunt.
Quote: "AI startup Hark advertised a $1 million giveaway, encouraging a download of Hark Pro and offering $30 in DoorDash credit as part of a promotion designed to showcase its agent's ability to order food."
Quote: "Hark acknowledged that the promotion strained its systems... 'we've made updates on how we handle peak surge requests.'"

Goldman Sachs

Description: Investment bank.
Why mentioned: Published an exclusive report naming power as the biggest risk to AI infrastructure spending.
Quote: "The explosive growth of AI is colliding with the glacial pace of building out new power infrastructure, prompting a new warning from Goldman Sachs that energy bottlenecks are set to stall the AI race."

Synchrony

Description: Consumer financial services company (newsletter sponsor).
Why mentioned: Positioning as an infrastructure player for agentic commerce standards and financing. (Sponsored content.)
Quote: "Synchrony is testing agent-ready financing experiences with select partners."

DoorDash

Description: Food delivery platform.
Why mentioned: The transaction rail for the Hark promotion.
Quote: "I did receive the promised $30 DoorDash credit, but the agent couldn't complete the order."

Target

Description: Retailer.
Why mentioned: The checkout where Muse got stuck and Dots routed via the saved-credentials app, illustrating the authentication problem.
Quote: "Muse attempted to complete the purchase but got stuck at Target's sign-in page."

Anthropic

Description: AI lab.
Why mentioned: Eric Schmidt's reported 1% stake, a signal of its valuation and insider exposure.
Quote: "Former Google CEO Eric Schmidt owns 1% of Anthropic, per Bloomberg."

SpaceX

Description: Aerospace company.
Why mentioned: Musk's announced share giveaway.
Quote: "Elon Musk says he will give one of his SpaceX shares to every 'kid' in America next year."

Intel

Description: Semiconductor company.
Why mentioned: Cited as a catalyst for the tech/AI stock rebound.
Quote: "The AI trade is all the way back as Intel helped tech stocks rally, per WSJ."

Amazon

Description: E-commerce and cloud giant.
Why mentioned: Its signature six-page memo culture is cited as degraded by AI-generated content.
Quote: "Amazon's famed six-page memo is no longer as useful because it's been turned into AI slop."

4. People Identified

Jason Tofsky

Description: Global head of digital infrastructure investment banking at Goldman Sachs.
Why mentioned: Source of the "forever bottleneck" framing for power.
Quote: "Power is probably the forever bottleneck."

Eric Schmidt

Description: Former Google CEO.
Why mentioned: Reported 1% holder of Anthropic.
Quote: "Former Google CEO Eric Schmidt owns 1% of Anthropic, per Bloomberg."

Elon Musk

Description: CEO of SpaceX, Tesla and other companies.
Why mentioned: Announced a SpaceX share giveaway to American kids.
Quote: "Elon Musk says he will give one of his SpaceX shares to every 'kid' in America next year."

Ina Fried

Description: Axios tech journalist.
Why mentioned: Hands-on tester of Dots, Muse and Hark; source of the practical findings.
Quote: "I decided to put two of the newest personal AI agents, OpenAI's Dots and Meta's Muse, through their paces."

Madison Mills

Description: Axios reporter.
Why mentioned: Author of the Goldman exclusive.
Quote: "Silicon Valley moves fast and breaks things, but the U.S. electrical grid moves slowly and requires permits."

5. Operating Insights

Design agent flows around existing authenticated state, not around full autonomy

Handing off to an app or session where credentials already live can beat trying to automate the entire checkout.
Quote: "Dots instead directed me to the Target app, where my account and payment information were already saved."

Load-test promotional launches for peak surge before you stunt

Hark's giveaway exposed capacity limits at the worst moment, creating public negative reviews from investors and founders.
Quote: "'We're so happy the majority of people had a great night with a free pizza,' a spokesperson said. 'Obviously, we hate to have let anyone down and we've made updates on how we handle peak surge requests.'"

Position agents as "research and legwork," with the human owning the decision

Users' comfort threshold sits at researching and comparing, not committing meaningful money. Product design and spending limits should reflect that.
Quote: "I'm not yet comfortable authorizing purchases of hundreds or thousands of dollars, or at a point of trusting them enough for consequential decisions."

6. Overlooked Insights

Agent self-correction as a trust-building feature

Muse proactively sent a correction after producing erroneous statistics, a behavior that could be a differentiator in trust-sensitive workflows.
Quote: "Muse later sent an unsolicited update correcting several factual errors."

Queue-speculation inflation in the power grid

The 2,600 GW of proposed projects exceeds existing grid capacity, but the caveat suggests the true constraint may be the quality and viability of the queue, not just its length. This matters for evaluating which energy developers can actually deliver.
Quote: "The U.S. has roughly 2,600 gigawatts of proposed power generation and storage projects awaiting grid connections, more than twice the grid's existing installed capacity, though many projects may never materialize."