• The Prohuman
  • Posts
  • What OpenAI's new science blog could reveal

What OpenAI's new science blog could reveal

Plus: Alibaba launches Qwen4-Max with stronger AI safety

In partnership with

Hello, Prohuman

Today, we will talk about these stories:

  • OpenAI launches a dedicated science blog

  • Alibaba says Qwen4-Max is faster and more efficient

  • Uber adds new safeguards for autonomous AI agents

The ice cream shop that makes money when it's cold

28 Wishes sells ice cream in Los Angeles. Below 70Β°F, sales fall about 20%. The weather is out of their hands. Rent isn't.

So the owners started putting about $20 a day into Kalshi weather markets, taking the cold side. The days that keep customers away now pay something back.

This is hedging. Big companies have done it for decades, buying protection against bad weather, fuel spikes and rising rates. It used to take a broker, a trading desk, and an order size no corner shop could meet.

Kalshi opens it up. Contracts on weather, fuel prices, inflation, tariffs and regulation, starting at a few dollars. Take a position on the outcome that would hurt you. If it hits, the payout softens it. If it doesn't, the contract expires and the good month was the point.

OpenAI wants to show how AI research actually works

image credits: OpenAI.

OpenAI is giving scientific research its own dedicated blog.

On November 15, the company announced a new publication where researchers can share AI-assisted discoveries, practical methods, and the limitations they encounter.

One example stands out. Thousands of GPT-Rosalind agents analyzed experimental data from millions of mouse brain cells in just 32 minutes, identifying biological mechanisms that scientists could investigate further in the lab.

I think the useful part of this announcement is OpenAI's commitment to publishing methods and explaining where they fall short. Researchers need enough detail to reproduce findings, especially when AI systems are producing mathematical proofs or suggesting experiments that could take months to validate.

If the blog consistently shares failed approaches and verification methods alongside successful results, it could become genuinely useful to scientists who want to apply these tools in their own work.

The question is how much of that process OpenAI will actually make public.

Alibaba is making AI safety part of its model pitch

image credits: Qwen

Alibaba has introduced Qwen4-Max, its latest flagship AI model, with a stronger focus on security and efficiency.

Announced on November 15, the model is designed for complex work across coding, finance, healthcare, and scientific research. Alibaba says it outperforms Qwen3.8-Max while using fewer tokens, potentially reducing the cost of running demanding tasks.

The security changes are worth watching. Alibaba says it added checks for code injection, memory-safety flaws, and broken access controls during training, while strengthening safeguards against cybercrime and other dangerous uses.

I think this reflects a growing expectation that powerful AI models should demonstrate basic security protections before businesses rely on them for sensitive work. A model that writes code or handles financial information needs to be evaluated for the risks it could introduce, alongside its accuracy and operating costs.

Alibaba's internal tests found no significant high-risk capabilities, although independent verification will matter as more developers begin using the model.

Will these security claims hold up in real-world deployments?

Uber wants to check what AI agents do before they act

image credits: Uber

Uber is introducing a security framework that checks whether AI agents should be allowed to perform specific actions, even when they already have permission to access the required systems.

The framework uses three layers of protection, covering agent registration before deployment, real-time checks during execution, and continuous monitoring afterward. It also records agent actions and tool calls so teams can review what happened and replay interactions when security policies change.

One detail matters here. Traditional access controls can confirm that an agent has permission to use a tool, but they cannot determine whether its actions make sense for the user's actual request.

I think Uber is addressing a practical problem companies will face as AI agents gain access to internal systems and start performing tasks independently. Giving an agent access to a database is relatively straightforward; making sure it uses that access appropriately across different situations requires more specific controls.

The challenge is cost and complexity, since each agent may need its own rules, testing, and monitoring.

Can companies maintain that level of oversight as they deploy hundreds of agents?

Prohuman team

Covers emerging technology, AI models, and the people building the next layer of the internet.

Founder

Writes about how new interfaces, reasoning models, and automation are reshaping human work.

Founder

Free Guides

Explore our free guides and products to get into AI and master it.

All of them are free to access and would stay free for you.

Feeling generous?

You know someone who loves breakthroughs as much as you do.

Share The Prohuman it’s how smart people stay one update ahead.