Day: August 14, 2025

Ukraine’s president, Volodymyr Zelensky, was not invited to the Trump-Putin summit in Anchorage, but 1,000 Ukrainian refugees in Alaska will be watching with trepidation.

• OpenAI first launched ChatGPT in 2022.
• OpenAI CEO Sam Altman now says the ChatBot boasts some 700 million weekly users.
• Here’s what you need to know about ChatGPT, and how it works.

Every time ChatGPT gets an upgrade, it’s news. But what exactly is ChatGPT?

ChatGPT is OpenAI’s flagship AI model. In August 2025, the company unveiled GPT-5, its latest iteration. OpenAI CEO Sam Altman says GPT-5 is its most advanced AI model yet, marked by increased general intelligence and enhanced usability, like a “real-time router” that selects the most appropriate model to handle each user request.

Here’s everything you need to know.

What is ChatGPT?

OpenAI is the leading AI startup. Its ultimate mission is to develop artificial general intelligence, or AGI, in a way that benefits humanity as a whole. AGI is a still-theoretical AI that reasons as well as humans.

The company first released ChatGPT in November 2022. It has yet to achieve AGI-level intelligence. For now, it’s a conversational chatbot that relies on a large language model to generate responses to questions. Some people use it much like they would Google Search. But it can also do deeper research, generate images and reports, write just about anything, code, and solve problems that involve quantitative reasoning.

Since its debut, the chatbot’s user base has exploded. OpenAI said in a blog post in August that its user base had reached 700 million weekly users.

How to use ChatGPT

ChatGPT is available online, and as an app available for both iOS and Android.

Users engage with it through conversation by simply typing in a prompt — an instruction for the chatbot. OpenAI also unveiled an “advanced voice mode” in 2024 — following a legal battle with Scarlett Johansson over the use of a voice that sounded too similar to hers — that lets users engage with the chatbot in natural, real-time conversations with the ability to sense emotions.

Since the release of ChatGPT, OpenAI has unveiled several different ChatGPT models — all of which can be used in conjunction with the chatbot. It has rolled out a series of reasoning models, for example, which are designed to think more deeply about problems. It also unveiled GPT-4.5, which Altman described on X as “the first model that feels like talking to a thoughtful person.”

Until the release of GPT-5, it was up to the user to understand which model was best for their needs. Now, GPT-5 can make that decision for them. That means, in essence, that the model is deciding how long it needs to think about a problem to get to the best answer.

ChatGPT also offers dozens of plug-ins to paying subscribers. An Expedia plug-in can help you book a trip, while one from OpenTable will nab you a dinner reservation. OpenAI has also launched Code Interpreter, a version of ChatGPT that can code and analyze data.

Despite the bot’s impressive capabilities, it remains imperfect. ChatGPT relies on available data for its responses, which means it can sometimes give misinformation. OpenAI has also been accused of stealing personal or copyrighted data to train ChatGPT. It has even encouraged students to cheat and plagiarize on their assignments.

How does ChatGPT work?

Chatbots like ChatGPT are powered by large amounts of data and computing techniques to make predictions and string words together meaningfully. They not only tap into a vast amount of vocabulary and information but also understand words in context. This helps them mimic speech patterns while dispatching encyclopedic knowledge.

When a user prompts a large language model, the query is broken into tokens — the smallest unit of text a model processes. For OpenAI’s models, they can be “as short as a single character or as long as a full word, depending on the language and context. Spaces, punctuation, and partial words all contribute to token counts,” according to OpenAI.

ChatGPT’s growing influence

Users have flocked to ChatGPT to improve their personal lives and boost productivity. The chatbot attracted 100 million users in its first five days on the market, a record at the time.

Some workers have used the AI chatbot to develop code, write real estate listings, and create lesson plans, while others have made teaching the best ways to use ChatGPT a career in itself.

Businesses, including consulting firms, are also scrambling to adopt AI. The popularity of ChatGPT crystallized the value of a conversational tool, McKinsey senior partner Delphine Zurkiya told Business Insider.

“There wasn’t a major shift in our strategy in the sense that we had already been developing a lot of tools internally. It’s just these tools now have become, we’ll say faster, in delivering value thanks to that natural user interface,” she said in regards to the firm’s internal chatbot, Lilli. Many consulting firms are also building similar tools for clients. KPMG, for example, has been collecting data on how its workers prompt AI, and used that information to build new tools — for itself and clients.

AI is also making waves in the legal world. Gibson Dunn is piloting ChatGPT Enterprise for its roughly 500 lawyers and staff. Judges, however, say they’ve seen an increase in fake legal citations due to lawyers relying too much on AI.

There is a slate of ChatGPT competitors that have also come out since its launch. Meta AI, built on its Llama 4 model, offers users an AI assistant that “gets to know” user preferences, remembers context, and is personalized. Anthropic’s Claude has become the leading AI assistant for coding. Elon Musk also built Grok, a chatbot that the company is training in line with Musk’s support for free speech. Google has Gemini, a multimodal model that CEO Sundar Pichai called “one of the biggest science and engineering efforts we’ve undertaken as a company.”

For OpenAI, which continues to unveil new models at a healthy clip, the chatbot is an eternal work in progress.

“There is no analogy for what we’re building,” Nick Turley, the company’s head of ChatGPT, said on a podcast in August.

• Infrastructure management startup NetBox Labs has raised $35 million, led by NGP Capital.
• Cofounder and CEO Kris Beevers said AI is fueling fresh demand for its services.
• NetBox Labs is the for-profit steward of the open-source NetBox platform.

NetBox Labs, an infrastructure management startup, raised $35 million in Series B funding in July, led by NGP Capital. It shared its pitch deck with Business Insider.

New York City-based NetBox Labs was cofounded in 2023 by CEO Kris Beevers as the for-profit steward of 10-year-old NetBox, an open-source network and infrastructure platform used by tens of thousands of companies globally.

NetBox Labs offers five products built on top of the open-source software with advanced features to make it easier to run, Beevers said, including additional security and support, automation tools, and an AI assistant.

It makes money by charging subscriptions based on the tools used and the size of a client’s infrastructure.

Sorenson Capital and Headline participated in the round, as did existing investors Flybridge Capital, Notable Capital, Mango Capital, Salesforce Ventures, Two Sigma Ventures, and IBM. The company had previously raised $20 million in Series A funding.

AI is fueling demand for infrastructure services, and NetBox Labs’ revenue is in the single-digit millions, Beevers said. It reinvests a “big chunk” back into the open-source tool.

The company has roughly 200 customers, including AI data center giant CoreWeave, for whom it helps manage data center configurations and design future facilities.

It also works with non-AI-focused companies, whose infrastructure has become more complex and dynamic amid ongoing digitization in recent years, Beevers said.

This includes departments of transportation across the country — managing toll booths, speed cameras, and more— as well as ARM, Cisco, Constant Contact, J.P. Morgan, Kaiser Permanente, and Riot Games.

The investment will be used to hire across engineering and sales. Beevers said the company will surpass 100 employees this year.

Beevers previously served as the cofounder and CEO of NS1, which spun out NetBox Labs after selling the other part of its business to IBM in 2023. Jeremy Stretch, the creator of NetBox, also serves as a cofounder at NetBox Labs.

Here’s a look at the pitch deck the startup used to raise its $35 million Series B. Certain slides have been edited and removed so that the deck can be shared publicly.

• The red-hot healthcare AI startup Abridge is looking for deals.
• Abridge has raised $700 million in the past 18 months and hit a $5.3 billion valuation in June.
• CEO Dr. Shiv Rao said the startup is considering various buys to bulk up its team and tech.

Abridge is on a hot streak. Now, as the healthcare startup competes in the AI talent wars, it’s preparing to go on a shopping spree.

Abridge, which builds AI-powered software to transcribe and summarize patient-doctor conversations, landed a $300 million round in June led by Andreessen Horowitz. It was its second mega-raise in four months; the company had raked in a $250 million round in February.

In 18 months, Abridge has raised $700 million in total and boosted its valuation from $850 million to $5.3 billion.

Abridge CEO and cofounder Dr. Shiv Rao has big plans for all of that capital, he told Business Insider.

“We’ve got a lot of money in the bank, and we want to spend 80% of it just going deeper, and expanding from connecting conversations to clinical notes into claims, clinical decision support, and care management,” he said.

With 80% of its cash reserved for doubling down on its tech, Abridge is earmarking the remaining 20% for potential acquisitions, Rao said.

Those deals are looking increasingly critical in the brutal, competitive landscape of ambient healthcare scribes. Perhaps the most gutting release is set to come from Epic, the electronic health record giant that first partnered with Abridge back in 2023 in a watershed moment for the startup. Epic is now gearing up to launch an AI scribe of its own, Politico reported this month. Abridge declined to comment on the reports. Epic didn’t respond to requests for comment.

Epic’s launch would follow healthcare startup Ambience, Abridge’s biggest direct competitor, raising its own mega-round this summer. Ambience’s $250 million Series C was co-led by Oak HC/FT and A16z, a longtime investor in Ambience that appears to be double-dipping in AI scribing with its growth investment in Abridge.

And, as Abridge works to expand beyond scribing to use AI for tasks like processing medical bills, it’s running up against private equity firms like New Mountain Capital pouring billions into their healthcare AI plays.

Abridge has remained a frontrunner as the pressure builds. It’s now working with over 150 large health systems, Rao said. The startup is also partnering closely with some of those health systems on its new products — Abridge said Tuesday it’s collaborating with Pittsburgh health system Highmark Health on tech to automate prior authorization requests.

And Abridge is focusing squarely on the patient-provider conversation as the starting point for each of its new tools or acquisition targets, Rao said.

“The last thing we want to do is to become a company that’s opening up a trench coat and selling you random things that have no coherence to our mission,” he said. “But if there are things that are absolutely on our road map, it would be smart for us to have open ears.”

Eyeing fresh talent and tech

As San Francisco-based Abridge has landed fundraise after fundraise, many startups hoping to get acquired by Abridge have entered Rao’s inbox.

While Rao said Abridge isn’t “in talks” with any particular company, it’s prepared to notch deals to grow faster. The startup hasn’t made any acquisitions since its 2018 founding.

“It feels like a lot of companies are asking if they can join us in some way. We need to be able to spend on things like that — data plays, ecosystem plays, and partnerships,” Rao said.

Acquiring top talent is Abridge’s biggest priority, Rao said, adding that the startup has been competing with AI giants like OpenAI and Anthropic in its recent hiring efforts.

The AI talent wars are raising the stakes for startups like Abridge. As Big Tech companies fight over top AI researchers, including by offering pay packages in the millions or tens of millions, startups are touting their mission-centric approaches to convince engineers to join their teams over tech giants.

Rao thinks Abridge can compete effectively for AI talent by offering hires the rare chance to build tools that matter, tech that actually improves people’s health.

“Finding ways to recruit more world-class talent as fast as possible is really, really important for us,” he said. “If this is the legacy you want to leave, if you want to be a part of a company where every single day you can feel really good that you improved patient care, then we’re going to resonate more than those horizontal technology companies.”

Abridge has about 330 employees, a number Rao is aiming to increase massively, especially in its engineering department, though he says the company won’t rely exclusively on M&A to do that.

Abridge builds its own large-language models that its widening suite of software sits upon. Those models make up its “contextual reasoning engine,” which automates clinical notetaking that can combine ambient scribing with relevant context from the patient’s existing health records and generate actionable outputs like medical orders and suggested billing codes. Rao said Abridge is considering buying data where necessary to continue training its models.

With a combination build-and-buy strategy, Abridge is moving further into revenue cycle management, the hottest ticket item in healthcare AI, since it aims to help health systems capture more revenue while saving time for doctors. Rao said Abridge is also working on areas like risk adjustment, the process of estimating a patient’s expected medical costs, that is critical for value-based care arrangements, and care coordination.

Abridge wants to dig deeper into clinical decision support, too, a field that many healthcare startups have stayed away from, as the tech often walks a thin line to avoid facing FDA regulations. Abridge first stepped into the space in October by partnering with medical insights company UpToDate to surface relevant clinical evidence in Abridge’s generated notes. Rao said he expects Abridge to share more information from the partnership later this year.

As Abridge looks to take over more tasks for doctors, the company is being deliberate about how and when it’ll meet “good friction” like FDA regulation, Rao said.

“As we move into higher-stakes workflows from a patient outcomes perspective, we have to be really, really responsible,” he said. “We try to be as transparent as we possibly can on how our models work and how we evaluate them. We’ll need to continue to be transparent as we get into those new spaces.”