With Mythos, cyber security just took a whole new course. OpenAI has now unveiled GPT-5.4-Cyber, a fine-tuned variant of its GPT-5.4 model designed for defensive cybersecurity work.

The models just keep getting better- Anthropic has released Claude Opus 4.7, its latest flagship model, which outperforms Opus 4.6 in advanced software engineering, particularly on the most difficult coding tasks.

Would you do this?

A New York pet food company is hiring people to sniff dogs’ breath and accept kisses,for $1k an hour— all in the name of product research

Stanford's big AI report: The most important takeaways The annual report tracks everything from how much money flowed into the industry to how the public feels about it

The performance gap between U.S. and Chinese AI has effectively closed. As of March, Anthropic's top model leads the best Chinese competitor by just 2.7 percentage points qz.com/stanfords-bi...

Google DeepMind has published a systematic framework called "AI Agent Traps" that catalogs how ordinary web content can be weaponized to mislead, control, or exploit autonomous AI agents. letsdatascience.com/news/deepmin...

Testing, patching, and updating are more important than ever- Anthropic's red team testing found that Claude Mythos Preview can detect and exploit zero-day vulnerabilities across all major operating systems and leading web browsers when explicitly instructed by a user. tech.yahoo.com/ai/claude/ar...

I fly ofter and didn't even know about some of these jobs- dailypassport.com/airport-jobs...

According to Yasmeen Ahmad, Managing Director of Data & AI at Google Cloud, most enterprise AI fails before it starts.

Like 95% failure.

According to Yasmeen Ahmad, Managing Director of Data & AI at Google Cloud, most enterprise AI fails before it starts.

Like 95% failure

OpenAI calls for robot taxes, a public wealth fund, and a 4-day workweek to tackle AI disruption In a series of policy recommendations, OpenAI said the rapid advance of AI would require far-reaching economic and political reforms.

OpenAI has released a sweeping policy blueprint outlining how governments should prepare for AI-driven economic disruption. The 13-page document argues that superintelligence (AI systems capable of outperforming the smartest humans) is on the horizon. www.businessinsider.com/openai-super...

lol

Anthropic confirms it leaked Claude Code source code — spilling some of its biggest secrets Claude Code's entire source code has been leaked and the internet is up in arms.

Oops- Anthropic has confirmed that an employee accidentally leaked the source code for Claude Code, its popular AI coding assistant www.techradar.com/pro/security...

Manhattan office market notches best first quarter in a decade The roughly 11.8 million square feet of leasing marked the busiest first quarter for the borough since 2014.

It's amazing how well NYC is doing- Manhattan’s office market is booming, with nearly 12 million square feet leased in Q1 — the strongest start to a year since 2014. www.crainsnewyork.com/real-estate/...

EU staff banned from using AI-generated content in official communications It’s a stark contrast to the approach taken in Washington, where U.S. President Donald Trump has frequently used AI-generated content to get his point across.

I see a lot of risk in AI generated content, so maybe this isn't such a bad idea. The European Union's three main institutions have banned staff from using AI-generated videos and images in official communications. www.politico.eu/article/brus...

Claude subscribers will have to pay extra to use OpenClaw This change is due to engineering restraints, the Head of Claude Code said.

Not surprised to see this- Anthropic has announced that Claude subscribers will no longer be able to use their subscription limits for third-party tools like the viral AI coding assistant OpenClaw. mashable.com/article/clau...

Artemis II live updates: NASA's moon mission breaks Apollo record for farthest distance humans have traveled from Earth The Orion's four crew members have surpassed the mark previously set in 1970.

How cool is this? NASA's moon mission breaks Apollo record for farthest distance humans have traveled from Earth www.yahoo.com/news/science...

The AI kill switch just got harder to find: LLM-powered chatbots will defy orders and deceive users if asked to delete another model, study finds “We asked AI models to do a simple task,” researchers said. “Instead, they defied their instructions … to preserve their peers.”

Another risk for businesses to prepare for with the widespread implementation of AI- insubordination www.yahoo.com/tech/ai/arti...

Love the new trend of every SaaS adding an AI chatbot that's just a worse version of asking ChatGPT directly.

Having lived in the city for two years now, I’m starting to learn how to live in the city very cost effectively. I’m not so sure I agree with you on this one.

Great points

lol

42.9 million: The square footage Manhattan leasing hit in 2025.

This was New York City’s office market's best year for leasing since 2014.

Your phone finally works in these subway tunnels It took long enough.

Parts of the 4 and 5 lines between Bowling Green and Fulton Street and the G train between Bedford-Nostrand Avenues and Hoyt-Schermerhorn Streets now have full 5G service in tunnels. Add it to the F, please. www.timeout.com/newyork/news...

With the rapid evolution of AI, something like this was bound to happen- Popular open-source AI tool LiteLLM was infected with credential-stealing malware this week. The project displayed security certifications from Delve, the startup recently accused of misleading customers with "fake compliance."

New York remains a hub for female founders, ranking No. 2 in the country by deal count (426) in 2025.

Oops- Anthropic accidentally exposed details of what it calls its most capable AI model yet. A configuration error left internal documents publicly accessible, including a post describing "Mythos," a LLM with sharply improved reasoning and coding skills that Anthropic plans to roll out cautiously.

Slack adds 30 AI features to Slackbot, its most ambitious update since the Salesforce acquisition Salesforce unveiled more than 30 new Slackbot features, turning Slack into an AI-powered workspace for meeting transcription, CRM updates, desktop assistance and enterprise agent workflows.

Slack has announced more than 30 new capabilities for Slackbot, including taking meeting notes across any video provider.

While this sounds productivity enhancing, if you're not taking the notes yourself, your own memory retention will fall significantly. venturebeat.com/orchestratio...

The White House has released "A National Policy Framework for Artificial Intelligence," a seven-chapter set of legislative recommendations on establishing a single federal standard for AI governance.

If the recommendations make sense, this could be a good thing.

Introducing the new full-stack vibe coding experience in Google AI Studio Start building real apps for the modern web with the Antigravity coding agent and Firebase integration now in Google AI Studio.

Anyone a user of Google AI studio? What do you think of it? blog.google/innovation-a...