
Welcome to OHMO AI. OpenAI wants ChatGPT to be where you find software, not one more app on the phone.
Trump and a large group of AI company leaders signed a voluntary accord at the White House, and Robinhood is giving all of its customers access this week to an AI agent that can carry out trades. And the government's own new chatbot began changing its answers while the President was still speaking.
In today's newsletter:
OpenAI launches Dots agents, shared Spaces and a marketplace aimed at the app store
Trump and the AI chiefs sign a voluntary accord to audit themselves
Robinhood's customers get access to AI trading agents starting this week
The government's America.gov chatbot rewrote its answers mid-launch
LATEST IN AI
OPENAI
OpenAI's Dots agents arrive with a marketplace aimed at the app store

Image source: OpenAI
OHMO AI: OpenAI used its Dev Day on Tuesday to launch Dots, personal AI agents, alongside a shared workspace called ChatGPT Spaces and an app marketplace with 30-plus partners signed up at launch. Put together, the announcements point at something bigger than a new feature: the app store model itself.
The details:
Dots are OpenAI's answer to Meta's Muse system, which is already popular, and the bet is that OpenAI's consumer lead can carry over into agents.
Spaces let coworkers, ChatGPT and their Dots draw on the same shared information, and they're for Pro, Business and Enterprise plans only, on the desktop app and the web before mobile.
Launch partners on the marketplace include Adobe, Figma, Salesforce, HubSpot, ServiceNow, Harvey, CrowdStrike and Palo Alto Networks.
OpenAI also announced new AI models at the same event, though the agents and the marketplace were the headline.
Why it matters: Salesforce, HubSpot and ServiceNow putting their apps on OpenAI's shelf means a customer's first stop can be ChatGPT rather than the vendor's own site, and no terms on whether OpenAI takes a cut came with it. Without terms, the marketplace is a claim about leverage rather than a revenue line. The role shift is already in the partner list: the companies that used to own the front door are now tenants behind someone else's.
PRESENTED BY TIGER DATA
You're Running Three Databases. You Only Need One.
Events go in a metrics store. Embeddings go in a vector database. Analytics get their own warehouse… Now you're running three systems, three sets of tooling, and pipelines to keep them in sync, all for one app.
TimescaleDB collapses that back into the Postgres you already run. Hypertables handle events at scale. pgvector and pgvectorscale handle embeddings. Continuous aggregates handle real-time analytics. It’s the same SQL, same tools, and only one system to operate. The sync jobs disappear. The drift disappears. The second and third databases disappear.
It's still Postgres, so nothing about your workflow changes except how much you have to maintain. Start on Tiger Cloud and get $1000 in credits.
WHITE HOUSE
Trump signs a "morally binding" AI accord with the industry's chiefs

Image source: CBS
OHMO AI: President Donald Trump said on Tuesday that he and a large group of AI company leaders signed a voluntary accord after a lunch at the White House, a meeting billed as a response to Americans' fears about the technology. Trump called the document "morally binding", while House Speaker Mike Johnson described it as a statement of principles that are "voluntary on behalf of the industry" and said the White House will guide how the industry develops. It landed as pressure for a slowdown in AI development grows louder.
The details:
Signatories included Anthropic CEO Dario Amodei, Google CEO Sundar Pichai, Meta CEO Mark Zuckerberg, OpenAI President Greg Brockman, Nvidia CEO Jensen Huang and Tesla and SpaceX chief Elon Musk.
The accord commits the companies to internal controls, an "independent external auditor" to assess whether those controls work, and a committee on each company's board to evaluate the auditors' reports.
Outside the White House afterwards, Amodei repeated his earlier concerns about AI and said the rules to address those risks are "still under discussion."
Amodei had a private dinner with the president on Sunday, after his absence from the state dinner for the Chinese president.
Why it matters: Every check in the accord reports inward. An outside auditor tests the controls, and its findings go to a committee of the company's own board, so the one body that can act on a failed audit is the company that failed it. That leaves each board as judge of its own auditor, and the White House, in Johnson's telling, as a guide rather than a referee.
ROBINHOOD
Robinhood hands roughly 29 million customers an AI agent that trades

Image source: OHMO AI
OHMO AI: Starting this week, all of Robinhood's customers will have access to AI trading agents, every account rather than a waiting list. The agents are built on models from OpenAI and Anthropic, and they're built to act, not only to answer.
The details:
Access reaches all of Robinhood's roughly 29 million customers, starting this week.
Customers can choose between various OpenAI and Anthropic models rather than being assigned one.
An agent can be told to carry out trades, do research, and build complex investment strategies.
Why it matters: The thing being handed over is the decision, not only the order. A customer's first new job is choosing between OpenAI and Anthropic before anything is bought, and the swap cuts the other way too: with both rivals a tap apart inside one app, Robinhood holds the switch and the model makers are the ones who can be replaced.
PRESENTED BY FRONT
A fast AI answer isn't always a coordinated one. When AI resolves B2B issues without looping in the teams responsible for them, confidence outpaces alignment. A new Harvard Business Review Analytic Services briefing paper, sponsored by Front, looks at closing that gap.
US GOVERNMENT
America.gov said Biden won and then changed its answer while Trump spoke

Image source: OHMO AI
OHMO AI: The US government launched America.gov on Tuesday, an official AI chatbot that says its answers are drawn from "official government sources". Asked who won the 2020 election, it said Joe Biden did, and on that and on the name of the department run out of the Pentagon it gave answers that contradict what Trump himself says. As Trump was still speaking at the launch event, the site started modifying its answers.
The details:
The White House fact sheet didn't say which AI company built it, but the tool runs at least partially on Google's Gemini platform.
Joe Gebbia, chief design officer of the National Design Studio, indicated that SpaceXAI's Grok is in there too, though that company hadn't publicly announced its involvement.
It has proved hard to jailbreak so far, and it carries Minecraft easter eggs: ask it about the game and it appears to have something like an existential crisis.
Why it matters: The answer on the 2020 election moved while the President was still at the podium, which means the official sources have an editor, and the editor was reachable inside the length of a speech. For anyone asking it about their benefits, the answer they get is whichever one survived the last edit.
AI GUIDES
Get a worked-through answer on hard problems by turning up Claude's effort
Ask Claude a hard question at the default settings and you get a quick, decent answer. Open the model menu next to the send button, raise effort to high and switch thinking on, and the same question gets a thorough answer with the reasoning laid out so you can check it. Two clicks, nothing to install, and you can drop back to low or medium for everyday chat so your usage limits stretch further.
Copy this, swap in the bracketed parts, and paste it in.
I need a careful, thorough answer, not a quick one. Here is the problem: [describe it in full, with the numbers, constraints, deadlines and what a good outcome looks like]. Work through it before answering, state any assumptions you make, and finish with a short summary of your answer and the one thing I should double-check.High effort and thinking give Claude room to reason through math, hard coding, planning and document analysis before it answers, and the expandable Thinking section shows you how it got there. Being specific about the problem still matters most, so fill the brackets fully.
COMMUNITY
Featured Reader
Every newsletter, we feature how a reader is using AI to work smarter, save time, or make life easier.
Today's reader is Thibault R. from Montpellier, France:
“Told everyone AI writes like a brochure. Still think that, mostly.
First week, September last year: asked Le Chat to write a client a note about my rates. Got three paragraphs of nothing. Deleted it. Point proved, I thought.
By the second month, different ask. Paste the whole email thread, then: what is this client asking me to decide, in one line. Then: answer it three ways, firm, neutral, soft, all saying the rate goes up in January. Sent the soft one. Client paid. Now it's a saved instruction in Le Chat, in the browser: my rates, my terms, never apologise. Every awkward reply goes through it. The URSSAF letters too, when they ask for the same attestation a third time.
Still won't let it write anything a client keeps. Only the replies I'd rather not.”
Want to be in the next newsletter? Tell us how you use AI here.
AI Meme

Go Deeper
AI Guides: Practical AI guides, prompts and tools.
Instagram: The AI page they pray you’ll never find.
Facebook: For the ones who never left Facebook.







