Anti-AI Bias Trick; OpenAI 🍽️ Personal Finance; Robots Work / Humans Slack

Psych! 'Monet-Gate' Turns AI Rage Into Performance Art

Conceptual artist SHL0MS posted an image from Claude Monet’s Water Lilies series and claimed it was AI-generated, asking X users to explain why the “AI image” was inferior. The responses came fast: emotionless, sloppy, bad reflections, weak composition. Then came the punchline: The image was a real Monet from around 1915.

Why it matters

The experiment exposed how quickly the label AI-generated can override actual judgment. In creative circles, “AI” has become less a tool category and more a tripwire.

The Deets

• SHL0MS framed the Monet image as machine-made and invited detailed criticism.
• Users attacked specific artistic choices, including depth, reflections and composition.
• The painting came from Monet’s Water Lilies collection.
• The stunt echoed research suggesting people may like AI art until they know it is AI.

Key takeaway

The backlash was less about brushstrokes than branding. Once people believed a machine made it, the art became guilty until proven Monet.

🧩 Jargon Buster - AI slop: A dismissive term for low-quality AI-generated content, often used broadly enough to hit things that are neither AI nor low-quality ... like this newsletter!

🧠 Power Plays

OpenAI Coming For Personal Finance

OpenAI rolled out a new personal finance experience inside ChatGPT, partnering with Plaid so users can connect bank, brokerage and bill data for personalized insights. The feature starts with U.S.-based Pro users and can be accessed through a finance sidebar or by tagging @finance in chats.

Why it matters

Money advice has long been split between expensive human advisers and clunky budgeting apps. ChatGPT now gets a shot at becoming the interface for spending, investing and bills, with Plaid playing the trust-and-security bouncer at the door.

The Deets

• Users can connect accounts from Chase, Schwab, Robinhood and more than 12,000 institutions.
• ChatGPT can analyze spending, investments and upcoming bills.
• It cannot move money, pay bills, make trades or file taxes.
• Intuit support is planned for tax estimates, credit-card approval odds and access to live experts.

Key takeaway

ChatGPT is moving from answering questions about money to sitting next to the money. That's powerful, useful and privacy-sensitive in equal measure.

🧩 Jargon Buster - Plaid: A financial data network that lets apps securely connect to users’ bank and investment accounts.

OpenAI’s Nation-Scale Onboarding

OpenAI is making Malta the first country-level ChatGPT Plus rollout, according to AI Secret. Citizens take an AI literacy course, then receive one free year of ChatGPT Plus. Officially, it is about education. Strategically, it gives OpenAI a small, tech-forward national test bed for everyday AI habits.

Why it matters

Malta’s population is under 600,000, which makes the program small enough to manage but large enough to study how students, workers, teachers and public institutions behave when one AI system becomes default infrastructure.

The Deets

• Citizens complete an AI literacy course before receiving Plus access.
• The rollout gives OpenAI public-sector credibility and broad user adoption.
• The playbook could be replicated with other governments.
• The real asset may be habit formation, not just subscriptions.

Key takeaway

The biggest platform wins are often quiet. Get people learning, working and asking inside one system, and the interface becomes the institution.

🧩 Jargon Buster - AI literacy: The ability to understand what AI tools can do, where they fail and how to use them responsibly.

🛠️ Tools & Products

Manus Turns A Cloud Computer Into Your Little Web-Crawling Intern

The Rundown AI shared a workflow for using Manus Cloud Computer as a private, always-on web crawler. Instead of burning agent tokens repeatedly checking the same site, users can set up a small script, schedule it with cron and log useful results to a CSV.

Why it matters

This is the kind of AI workflow that feels less flashy and more useful. Agents do not need to think every second. Sometimes they just need to set up boring automation that runs cheaply and reliably.

The Deets

• Users create a Cloud Computer inside the Manus desktop app.
• Manus can inspect a website, identify the best monitoring method and write a crawler script.
• The task can run twice daily, log every run and save only matching results.
• Users can later ask Manus to check the collected data from a phone or laptop.

Key takeaway

The agent does the setup. The cloud box does the grunt work. That division of labor is where a lot of practical AI automation is heading.

🧩 Jargon Buster - Cron: A scheduler that runs scripts automatically at set times, like twice a day or every Monday morning.

💰 Funding & Startups

A $1.3M Codex Bill Shows The Cost Of Agent Armies

OpenClaw creator Peter Steinberger’s team reportedly showed a bill of $1.3M in 30 days for about 100 Codex agents, or roughly $13,000 per agent per month. The eyebrow-raising part was not just the cost... the agents were handling messy operational work around software, including issue triage, bug cleanup, regressions, reviews and support noise.

Why it matters

Software jobs include a lot of glue work that rarely appears in glamorous coding demos. If agent costs fall, companies may aim them at the hidden labor layer around engineering, where duplicated tickets, stale pull requests and failing tests quietly eat entire teams.

The Deets

• The reported bill covered about 100 Codex agents over 30 days.
• Much of the work focused on operations around the OpenClaw community.
• Tasks included bug cleanup, issue triage, reviews, regressions and support handling.
• The cost looks wild for an individual, but potentially cheap for companies replacing repetitive coordination work.

Key takeaway

The agent economy may first eat the inbox, the bug tracker and the support queue before it eats the senior engineer.

🧩 Jargon Buster - Agent: An AI system that can take multi-step actions toward a goal, such as writing code, checking tests or managing tickets.

🤖 Research & Models

Figure’s Robots Sort While Humans Sleep

Figure’s factory livestream passed the 100-hour mark, with its humanoid robots sorting more than 80,000 packages at a steady 2.9-second cadence. The reported pace landed around 1,240 to 1,250 packages per hour, using battery shifts and wireless charging docks.

Why it matters

Warehouse sorting is a brutal proving ground because speed, consistency and fatigue matter. Figure’s livestream is turning humanoid robotics from polished demo theater into an endurance test with real labor implications.

The Deets

• Figure’s fleet ran past 100 hours in an unedited factory livestream.
• Robots sorted more than 80,000 packages.
• The system held a roughly 2.9-second rhythm.
• CEO Brett Adcock addressed questions about teleoperation, supply-chain independence and future Figure 4 designs.

Key takeaway

Sorting work is becoming a robotics beachhead because the job rewards exactly what machines are getting better at: repetition, endurance and consistency.

🧩 Jargon Buster - Teleoperation: Remote human control of a robot, often used during testing or for difficult edge cases.

⚡ Quick Hits

Anthropic expanded its partnership with PwC, bringing Claude Code and Cowork into consulting, deals and enterprise transformation work.

Anthropic also formed a $200M partnership with the Gates Foundation to use Claude in vaccine screening, disease forecasting and K-12 tutoring in developing nations.

OpenAI reportedly acquired Weights.gg, a voice-cloning social network with celebrity replicas, with six staffers joining the company.

Pope Leo XIV established a Vatican AI advisory body and signed his first encyclical, expected to frame AI as this generation’s Industrial Revolution.

YouTube is opening its AI likeness-detection tool to creators 18 and older, giving more users a way to find and remove deepfake clones.

Nvidia nearly doubled its CoreWeave stake and added Coherent shares, tightening its role in the AI infrastructure stack.

Microsoft introduced a WinUI agent plugin for GitHub Copilot and Claude Code to help AI agents build cleaner native Windows apps.

Vercel introduced Zero, an experimental systems language designed so AI agents can read compiler feedback and repair native programs more easily.

Gallup found that 70% of Americans oppose nearby data centers, making the infrastructure less popular than local nuclear power plants in the survey.

Today’s Sources: The Internet, The Rundown AI, AI Secret, Robotics Herald