PRODUCTS

What ChatGPT will do with its Codex superpowers

OpenAI is collapsing the wall between ChatGPT and Codex, folding them into a single platform so users no longer have to toggle between apps, and opening Codex to far more people than the coders who first relied on it.

Instead of making you choose between using the fast-growing Codex agent and ChatGPT, which has become the company’s biggest brand used by nearly a billion users, the company decided to take all of Codex and put it inside ChatGPT, Alex Embiricos, head of enterprise product, explained in a Codex workshop for press in New York City. 

The result: easier access to Codex’s powerful capabilities.

"It's been really fun working with software developers, because they're an audience that wants to try a lot of new things," said Embiricos. "We're now at the second phase, where we have these incredibly useful agents that are actually useful to do anything you can do on your computer. And now our goal is to bring this capability to everyone." 

This update will also improve interoperability, Embiricos explained, as users on a computer or phone will have access to the same Codex capabilities directly within ChatGPT. 

Also this week, OpenAI launched six new role-specific plugins that target roles beyond traditional software development, such as creative production, sales, and public equity investing, highlighting the push to bring Codex to a broader audience. 

"For us, this is really core to our mission. We really care about making sure that the best technology is broadly available to everyone," said Embiricos. 

If, like me, you've been skeptical about what Codex can actually do for non-coders, the workshop I attended shed some light on a few ways everyday users can put it to work. Here's what I found genuinely useful:

• Sending messages: By connecting your email, Slack, or another messaging platform, you can describe what you want to say, and Codex will draft it, locate the recipient's information, and send it with a single approval from you.

• Calendar briefings: Connect your calendar via a plugin, and you can ask Codex questions about upcoming events, or have it surface specific information on demand.

• Daily automations: Once you've combined the right plugins for a given workflow, you can instruct Codex to run it automatically at whatever cadence works best for you. For example: Every morning at 8 AM, give me a summary of today's meetings and the most important action items from my inbox.

Tools like Codex and its archrival Claude Code have built their reputations on software development, and that association runs deep. Rebranding Codex to appeal to a broader audience would have been an uphill battle, especially given how intimidating the interface can feel at first glance. Folding it into ChatGPT, a product so many people already know and use daily, was the smarter move. The bigger challenge now is helping people understand just how much these tools are actually capable of. For those willing to tinker and embrace the learning curve, this merger stands to unlock a whole new level of productivity. But you have to be willing to change your mental model of ChatGPT from just something you chat with to something that accomplishes tasks for you. 

TOGETHER WITH CDATA

Single-source MCP servers fail 1 in 4 prompts

The failure patterns are consistent: schema mapping, date logic, multi-filter conditions, and write validations — the exact operations enterprise workflows depend on. Layer in the absence of centralized governance controls, and a single-source approach to MCP integration becomes a nonstarter at scale.

CData Connect AI resolves all four failure patterns, delivering a 98.5% accuracy rate. Governance is built into the architecture: per-user authentication, downscoped toolsets and datasets, and query-level audit trails.

See how Connect AI handles what single-source servers can't. →

ENTERPRISE

The tokenmaxxing era is over before it started

AI customers may be starting to pinch their pennies, and tech giants are taking notice. 

At both Microsoft Build and Snowflake Summit this week, efficiency stood out as a prevailing theme in the announcements of these enterprise tech giants. It may signal that the compute costs that are crunching AI builders are starting to add up, and flagrant spending fueled by sky-high expectations may be starting to come back to earth. 

"I think if you read about OpenClaw's founder, Peter Steinberger, and how many millions of dollars worth of tokens that he's using, it doesn't necessarily correlate to an output," Rob Ferguson, VP of technology and strategy at Fireworks AI, told The Deep View this week. "People are starting to really think about what the outputs of their AI are." 

In short, the era of "tokenmaxxing" may be over. Or, at least, the definition is changing, said Ferguson. Rather than focusing on eating up as many tokens as their competitors, enterprises are starting to think about how to squeeze as much as they can out of the tokens they use. 

Several of the product releases in San Francisco this week back up that shift: 

• Snowflake's new Cortex Training system, which allows enterprises to customize open-weight foundation models, is marketed specifically as being faster and less expensive. Additionally, Snowflake's new Adaptive Compute addresses cost efficiency at the infrastructure level by automatically calculating the best use of compute and software resources in real time. 

• Microsoft's new models also reflect a desire for efficiency, with its first reasoning model sitting at 35 billion parameters (compared to the latest trillion-parameter models that OpenAI and Anthropic offer) and built specifically for efficiency and low-token cost. 

• The company is even targeting efficiency on the hardware side, debuting both the Surface Laptop Ultra and the Surface RTX Spark Dev Box, which can run powerful models locally and drastically reduce token costs. Jatinder Mann, partner director of product management at Microsoft, told The Deep View that these devices aim to provide "unmetered intelligence," reducing cloud costs by enabling local models to handle routine tasks. "There are a lot of routine things that don't necessarily need a cloud model," Mann said.  

The next step enterprises should take is questioning whether a task requires AI at all, Raj Ramanujam, VP of Global Alliances and Cloud at Dynatrace, told The Deep View. Every agentic task, every prompt, every tool call racks up the bill. It's why every potential AI implementation should start with a "problem statement," he said, identifying exactly what challenge they're trying to solve or task they'd like to automate. 

"There are some things that you can automate without touching AI in the normal course of how you program it," said Ramanujam.

The tokenmaxxing fad has grown to the point where memes are going viral about companies burning tokens on agents who write poetry and send motivational messages. But what goes up must come down. Seeking to cement their statuses as AI-first, many companies felt the pressure to go all-in on the tech without considering the costs, racking up massive bills with models running in the cloud, and those bills have clearly started to sting. And with the ROI equation still unanswered, some enterprises may be feeling uneasy about their AI strategies. If these announcements signal anything, it's that big tech firms know they, too, need to lean into efficiency, rather than pressuring their customers to use up as many tokens as possible (looking at you, Jensen). 

TOGETHER WITH NICE

AI that performs at enterprise scale

Enterprise scale isn’t a claim. It’s what customers deliver with NiCE every day. Billions of journeys.

Every channel. Every interaction. Real AI-powered outcomes that help leading brands deliver smarter, faster customer experiences at global scale.

See why the world’s leading brands trust NiCE to power AI-driven customer experiences at enterprise scale.

RESEARCH

Why Anthropic's RSI warning rings hollow

AI developers have long been on a quest to make models that can train themselves. Anthropic thinks we're getting closer. 

On Thursday, the Anthropic Institute published findings on its progress towards recursive self-improvement (RSI), which involves models that can autonomously design and develop their successors. Though the company indicated that the tech is not yet at this point, nor is it inevitable, its research shows that "it could come sooner than most institutions are prepared for." 

If AI gets to the point where it is building itself, that step-change in development makes securing and monitoring the tech all the more important, Anthropic said. 

"AI that can build itself would be a major development in the history of technology … But full recursive self-improvement also might increase the risks of humans losing control over AI systems," Anthropic said in its blog post. 

As it stands, AI has already completely transformed software development: 

• At Anthropic alone, more than 80% of code merged into the company's codebase is now written by Claude, and engineers are shipping 8x as much code as they did in 2024. 

• Claude can now handle open-ended engineering problems, rather than just specific tasks, with far more accuracy, scoring a 76% success rate on these jobs this May, up 50 percentage points in just six months. 

• And going head-to-head with humans on research judgment, Anthropic's best models now suggest better next steps than humans 64% of the time. 

These growth areas mean that the role of humans is changing: rather than solving problems themselves, human judgment is devoted to deciding which problems are most important to solve. 

With the current trajectory of improvement, Anthropic outlined three potential scenarios. The first is that AI improvement stalls, but the capabilities of these models are spread more widely; the second is that AI continues to improve, but humans remain in the driver's seat; and the third is that RSI is achieved, humans play a "substantially diminished role" in creating them, and the skills of automated self-improvement are then transferred to other domains. The latter is how you get a future where "human labor stops being competitive," as Anthropic noted in its statement. 

Though Anthropic couldn't predict what this future would look like — or even if RSI will happen — the institute suggested that having the "option to slow or temporarily pause frontier AI development" could give governments, alignment research and societal structures enough time to catch up. That massive an initiative, however, would require agreement from multiple major labs in several countries. 

"We don’t have that long," the blog post reads. "A unilateral pause by one lab, by contrast, is achievable immediately, but accomplishes much less: it would change who the front-runner is, but it would not create the wider deliberative process that is currently missing."

As alarming as it is to consider AI that is capable of improving itself beyond human control, getting the industry to cooperate on safety is probably wishful thinking. Any voices begging the industry to slow the march into the unknown will likely be overruled by two camps: those who think RSI will usher in an AI-powered utopia, and those who don't believe in RSI at all. With trillions of dollars at stake, there isn't a financial incentive to stop and consider exactly how self-improving AI could fundamentally shift our society. It's a truth that Anthropic itself is aware of, given that in the last week alone, it released its latest powerful AI model, raised $65 billion, and filed for an IPO. And of course, as the blog post notes, Anthropic won't slow down if others don't, as that would simply give less benevolent actors time to advance rapidly without competitive pressures.

LINKS

• Airbnb CEO Brian Chesky plans new AI lab

• Robotics firm Generalist raises $400 million at $2 billion valuation

• Google debuts research tech to passively track heart rate via facial video

• Anthropic reportedly embeds a dozen forward deployed engineers in NSA

• House unveils draft legislation that would preempt state AI laws

• Canada invests $500 million in AI strategy to create 250,000 jobs

• ChatGPT Dreaming: OpenAI has rolled out a more capable and scalable system for synthesizing memory for Plus and Pro users in the US.

• Gemma 4 12B: A lightweight, encoder-free multimodal model by Google. 

• Poke: Apple has approved the agentic AI startup's tech in its Messages and Business platform. 

• Nemotron 3 Ultra: Nvidia's 550 billion parameter open model built for long-running agents.

• Meta: AI Research Scientist - Meta Superintelligence Labs

• Anthropic: Research Engineer, Safeguards Labs

• Dolby Laboratories: Senior Staff AI Researcher

• Nvidia: Developer Advocate – Agentic AI

GAMES

POLL RESULTS

Are you concerned about potential security vulnerabilities caused by vibe coding?

Yes (63%)
Somewhat (25%)
No (6%)
Other (6%)