Latest AI News
Ecolab Acquires CoolIT for $4.75bn to Strengthen AI Data Centre Cooling Business
With CoolIT’s expertise in liquid cooling, Ecolab aims to provide comprehensive solutions across the AI value chain, targeting $4 billion in annual revenue from its Global High Tech business by 2030.
View
Amazon will stop accepting new customers for Mechanical Turk
These may be the last days of Amazon’s Mechanical Turk. An announcement onthe Mechanical Turk websitesays that on July 30, 2026, the crowdsourcing service will close to new customers.Amazon Web Services saysthe decision was made after “careful consideration,” adding, “Existing customers can continue to use the service as normal. AWS continues to invest in security and availability improvements for Mechanical Turk, but we do not plan to introduce new features.” In other words, Amazon isn’t completely pulling the plug, but the service is very much on life support. First launched in 2005, Mechanical Turk was a marketplace where people were paid tiny amounts to perform simple tasks that resisted full automation — things like completing CAPTCHA challenges or identifying the basic sentiment in a sentence. In its heyday, the service was at the center ofdebates around the ethics of crowdsourced labor, and it evenplayed a small rolein the early stages of the Facebook-Cambridge Analytica scandal. Beginning in 2018, Amazon also beganbilling it as a way for companies to annotate datato train neural networks as part of its SageMaker AI service. Less overtly, Mechanical Turk has also been described asthe hidden enablerfor companies taking afake-it-till-you-make-itapproach to AI, where products marketed as Ai are actually being performed by the Mechanical Turk workforce — all the more fitting sincethe original Mechanical Turkwas itself a hoax, with a hidden human chess player pretending to be a chess-playing machine Over time, the relationship between Mechanical Turk and AI models grew even more complicated. In asnake-eating-its-own-tail irony, a 2023 analysis found that between 33% and 46% of workers on the platform were using large language models to complete their tasks,raising questions about the reliability of data annotated on the platformand also about whether humans needed to be in the loop at all. This week, after Amazon’s decision became public, one Reddit user suggested the platform died “years ago,” with workers and researchers abandoning it due to bots and fraud. The userpredicted, “Someone at Amazon is going to decide keeping the Mturk servers running is a waste of time and resources and pull the plug entirely.”
View
This Ex-Google Engineer Built a $3.5 Bn Startup to ‘Fix’ the Software Supply Chain
Raising more than $800 million, Dan Lorenc believes the biggest security problem in software lies in how enterprises consume open-source packages.
View
New Google commercial imagines a Declaration of Independence written with help from AI
Two hundred and fifty years after the signing of the Declaration of Independence,a new commercial from Google asks: What if the Founding Fathers had access to Google Workspace? With the tagline “Group project, but make it 1776,” the ad depicts a largely unseen Thomas Jefferson mid-draft when he gets a nagging text from Ben Franklin, leading to a very Google-centric collaboration process. Edits are suggested in Google Docs, a meeting gets scheduled in Google Calendar and conducted remotely via Google Meet (with every single attendee apparently turning their camera off?), then the whole thing is finalized with e-signatures; cue the fireworks. Of course, since this is an ad from a tech company in the year 2026, AI has a role to play. The fictionalized founders use Google’s “help me visualize” AI tool to try out different animals on the national seal, Gemini takes notes on the meeting, and the founders also ask the chatbot for advice before declining King George III’s document access request. The whole thing is very tongue-in-cheek (at one point, Sam Adams asks, “Can we settle this over beers?”), and the AI evangelism is relatively discreet when compared tomany other recent ads. And unlikethat infamous Google commercialin which a father uses Gemini to write a fan letter for his daughter, this one shies away from any suggestion that the actual text of the Declaration of Independence would be improved with AI. Perhaps the most AI-forward element of the ad is the footage itself, which to my eye has the uncanny glow of AI-generated video. While viewer comments onYouTubeandInstagramappear to be mostly positive, you may not be surprised to learn that the response on Bluesky has beenfar more critical. Posters declared the commercial “cringey” and “stunningly tone deaf,” and the AI angle was the biggest target — even as many users,including historian Angus Johnston, noted that it’s “amazing how little of this is actually AI.” “Even in a corny fantasy joke, it’s impossible to make the case that AI is a useful tool for political organizing, writing, or human collaboration,” Johnston said.
View
What is Mistral AI? Everything to know about the OpenAI competitor
Following the Trump directive that led Anthropic topull its latest AI models offlineand growing calls forsovereign tech that reduces reliance on the U.S.,Mistral AIhas been caught in a whirlwind of attention. But the French AI darling is often misunderstood, and the fact that it develops large language models (LLMs) has muddied the picture. Anyone who judges Mistral by how close it is to becoming ‘the OpenAI from Europe’ is in for disappointment. Its chat and agent Vibe, formerlyLe Chat, only has an ounce of ChatGPT’s brand recognition, and Claude is more popular than Mistral’s modelseven among founders based at Station F, Paris’ startup campus. On the other hand, casual observers tend to miss that the French decacorn is following the Palantir playbook, with forward-deployed engineers that help governments and large corporations adopt AI and tailor it for their use cases. This approach is also better suited for Mistral’s means. While the company is rumored to be raisingsome $3.5 billion at a $23.15 billion valuation, nearly doubling its current valuation, that’s still far less than U.S. frontier labs. But its revenues have also ramped up; in February, it disclosed that its annual recurring revenue was nowabove $400 million, up from $20 million just one year earlier, and claimed it was on track to surpass $1 billion in ARR this year. This has helped Mistral gain a seat at the table in places like Davos, and even in rooms where tech CEOs have a hard time getting their message across, such asthe French Parliament. Mistral CEO Arthur Mensch has becomea public ambassador for a certain vision of AI, but he still has some evangelizing to do when it comes to explaining his own company. In a lengthyLinkedIn post, Mensch broke down what the Paris-based company has been doing “for a living” — deploying its models and agent platform on the infrastructure of its Enterprise customers, and helping them build custom models withForge, a platform that lets them use their own data for training. However, misunderstandings and bigger hopes around Mistral don’t stem out of thin air. Named after a wind, the company pursues a grand vision. “We exist to make sure that everyone gets access to the best AI systems, outside of centralized control exercised by states or corporations that feel the need to control in-fine deployment of AI,” Mensch wrote. This vision means that Mistral is looking beyond the enterprise. It also aims to keep on making big investments into research to keep up with foundational AI rivals — and Mensch’s post also covered where he thinks the company stands in that regard. “Today, we do not yet own the best language models, but we’ve constantly reduced that gap. We have a very exciting model to come this summer – it will be open-weight, and we’re opening early access to it in July. In domains that are less compute bound, e.g. voice, vision and document processing, we have state-of-the-art solutions,” Mensch claimed. Mistral’s upcoming model has already generatedsome buzz on X, where Mensch and Mistral backer Marc Andreessen haveengaged with jokesand amplified memes on what we now know won’t be called “Le Chaton Fat.” That’s another sign that the world — especially “the rest of the world” — is keeping an eye out for whatever Mistral has in its bag. The most interesting part may be happening behind the scenes. Earlier this year, Mistralacquired infrastructure startup Koyebto further boost its plans to build “a true AI cloud. The company also announced a€4 billion investment strategy(around $4.56 billion) to build data centers in France and Sweden — and the sovereignty undertones are never very far. “We’re building under the premise that AI technology is a commodity technology that every organization needs a secured and affordable supply of,” Mensch wrote. If you are curious to learn more, keep on reading. Mistral’s three founders share a background in AI research at major U.S. tech companies that have operations in Paris. Before becoming Mistral’s CEO, Mensch used to work at Google’s DeepMind; CTO Timothée Lacroix and chief scientist officer Guillaume Lample are former Meta staffers. Mistral also granted the title of co-founding advisers to the cofounders of health insurance startupAlan, Charles Gorintin andJean-Charles Samuelian-Werve(also a board member). In addition, it recently appointed three new executives to support its growth: Johan Bergqvist as Chief Financial Officer, Brian Hall as Chief Marketing Officer and Kamal Brar as SVP, Partners & Alliances. Mistral has developed abroad suite of modelsranging from LLMs to multimodal, reasoning, audio andOCRmodels. Not all of its models emphasize size; there’s the tellingly named Mistral Small 4 and “Les Ministraux,” a family of modelsoptimized for edge devicessuch as phones. Some are open weights, and it alsomade code agent Leanstral open source. In 2024, Mistralsigned a deal with Microsoftthat included a €15 million investment and a strategic partnership for distributing the French company’s AI models through Microsoft’s Azure platform. In May 2025, Mistral said it would participate in the creation ofan AI Campus in the Paris region, as part of a joint venture with UAE investment firm MGX, NVIDIA, and France’s state-owned investment bankBpifrance. In June 2025, Mistral said it would launch a European platform dedicated to AI and powered by Nvidia processors,Mistral Compute, in 2026. The initiative washailed as “historic”by France’s president, Emmanuel Macron, who shared the stage with Mensch and Nvidia CEO Jensen Huang at the VivaTech conference shortly after the announcement. In July 2025, Mistral launchedAI for Citizens, an initiative that the company claimed could “help States and public institutions strategically harness AI for their people by transforming public services.” In September 2025, Mistral and chip company ASMLstruck a partnership“to explore the use of AI models across ASML’s product portfolio as well as research, development and operations.” Mistral also secured strategic partnerships with the likes ofAccenture,press agency Agence France-Presse, France’sarmyandjob agency,Luxembourg,shipping giant CMA, German defense tech startupHelsing,IBM,Orange, andStellantis. Most of Mistral AI’s funding to date wasdebt financing, but the company has also raised several venture funding rounds, with a grand total around $4 billion, according toCrunchbase. In June 2023, just one month after being founded, Mistral AI raised arecord $113 million seed roundled by Lightspeed Venture Partners. Sources at the time said the seed round,Europe’s largest ever, valued the startup at $260 million. Other investors in that round included Bpifrance, Eric Schmidt, Exor Ventures, First Minute Capital, Headline, JCDecaux Holding, La Famiglia, LocalGlobe, Motier Ventures, Rodolphe Saadé, Sofina, and Xavier Niel. Six months later, Mistral closed a€385 million Series A($415 million at the time), at a reported valuation of $2 billion. The round was led by Andreessen Horowitz (a16z) and saw participation from Lightspeed, as well as BNP Paribas, CMA-CGM, Conviction, Elad Gil, General Catalyst, and Salesforce. Microsoft’s$16.3 million convertible investmentin Mistral as part of a partnership announced in February 2024 was presented as a Series A extension, implying an unchanged valuation. In June 2024, Mistral raised€600 million (about $640 million) in a mix of equity and debt. Thelong-rumored roundwas led by General Catalyst at a $6 billion valuation, with notable investors including Cisco, IBM, Nvidia, and Samsung Venture Investment Corporation participating. In September 2025, Mistral closed a €1.7 billion Series C round (about $2 billion) led by ASML at a €11.7 billion valuation (approximately $13.8 billion), with participation from existing backers DST Global, a16z, Bpifrance, General Catalyst, Index Ventures, Lightspeed, and Nvidia. In addition toinfrastructure startup Koyeb, Mistral has also boughtEmmi, an Austrian startup focusing on physics AI, with the ambition to better support industrial enterprises in their AI transformation. While Mistral has yet to design its own chips, Menschisn’t ruling it out. “Owning the chips may come, I think it should come at some point, but for now we are relying on Nvidia, which is a great partner to us, and we’re testing a few things here and there,” he told CNBC. Mistral is “not for sale,”Mensch said in January 2025 at the World Economic Forum in Davos. “Of course, [an IPO is] the plan.” This makes sense, given how much the startup has raised so far: Even a sale to arumored prospective buyer like Applemay not provide high enough multiples for its investors, not to mention sovereignty concerns depending on the acquirer. This story was originally published on February 28, 2025, and will be regularly updated.
View
Alibaba reportedly bans employees from using Claude Code
China’s Alibaba will ban employees from using Anthropic’s programming tool Claude Code, starting on July 10, according tomultiplereports. Anthropic already prohibits Chinese companies, as well as foreign entities owned by those companies, from using its models. The company has reportedly beenworking to close loopholesthat allow Chinese users to access Claude. According toa recent Reddit post, some of that loophole-closing involved a version of Claude Code that could secretly identify Chinese users. Anthropic’s Thariq Shihiparsaid in a post on Xthat this was “an experiment we launched in March that was meant to prevent account abuse from unauthorized resellers and protect against distillation.” (Distillationis a practice where AI models are trained on the outputs of other models.) “The team has landed stronger mitigations since then and we’ve actually been meaning to take this down for a while,” Shihipar said. Nonetheless, Alibaba has reportedly classified Claude Code as high-risk software and is instructing employees to use the company’s own Qoder tool instead.
View
Midjourney wants Hollywood studios to reveal the details of their AI usage
As part of an ongoing legal dispute with three Hollywood studios, AI startup Midjourney is seeking to compel those studios to reveal how they use AI themselves. Disney and Universal sued Midjourney for alleged copyright infringementlast year, noting that the startup’s image-generation models could create images of characters, such as Bart Simpson and Darth Vader, who are owned by the studios. A few months later,Warner Bros. sued Midjourneyas well. The startup argues that training its AI models on images of copyrighted characters is permitted under fair use. The current dispute revolves around the documentation the studios will need to produce during the discovery process. A judge previously ruled that the studios would indeed have to provide information about their generative AI usage – but only when it led to “consumer-facing” videos and images. Inits latest filing, Midjourney seeks to overturn that limitation, arguing that it “unfairly” allows the studios “to cherry-pick only those documents they believe support their market harm claims while depriving Midjourney of documents that would support its defenses.” Midjourney goes on to claim that the “documents [the studios] are withholding are precisely those that would reveal whether, behind closed doors, they are doing exactly what they are suing Midjourney for doing.” For example, the startup says that if the studios are developing image-generating AI models “for internal use in storyboarding or ideating content for film or TV, that evidence would equally demonstrate that it is an industry custom, even among the studios themselves, to download and train AI on unlicensed copyrighted content.” In the filing, the startup also argues that the studios should reveal all the prompts they used in Midjourney, as well as the resulting outputs, not just the prompts that produced the allegedly infringing images. The studios’ lead attorneyDavid Singer previously claimed Midjourney was seeking this documentationas part of a “fishing expedition.” He also said the studios “do not seek to stop AI technology or even shut down Midjourney’s business,” but rather “simply want Midjourney to stop copying their movies and TV shows and to stop distributing, publicly displaying, publicly performing, and creating derivative works that include copies of [their] famous characters without authorization.”
View
In Swiggy's Developer Community, Some are Planning Parties, Other are Prepping Meals
Swiggy is enabling developers to build health assistants, accessibility tools, and conversational shopping agents.
View
The only AI glossary you’ll need this year
Artificial intelligence is rewriting the world, and simultaneously inventing a whole new language to describe how it’s doing it. Sit in on any product meeting, pitch, or panel these days, and you’ll hear people toss around LLMs, RAG, RLHF, and a dozen other terms that can make even very smart people in the tech world feel a little insecure. This glossary is our attempt to fix that: pain-English definitions of the AI terms you’re most likely to actually run into, whether you’re building with this stuff, investing in it, or just trying to keep up by reading TechCrunch or listening to related podcasts. We update it regularly as the field evolves, so consider it a living document, much like the AI systems it describes. Artificial general intelligence, or AGI, is a nebulous term. But it generally refers to AI that’s more capable than the average human at many, if not most, tasks. OpenAI CEO Sam Altman once described AGI as the “equivalent of a median human that you couldhire as a co-worker.” Meanwhile,OpenAI’s charterdefines AGI as “highly autonomous systems that outperform humans at most economically valuable work.” Google DeepMind’s understanding differs slightly from these two definitions; the lab views AGI as “AI that’s at least as capable as humans at most cognitive tasks.” Confused? Not to worry —so are experts at the forefront of AI research. An AI agent refers to a tool that uses AI technologies to perform a series of tasks on your behalf — beyond what a more basic AI chatbot could do — such as filing expenses, booking tickets or a table at a restaurant, or even writing and maintaining code. However, as we’veexplained before, there are lots of moving pieces in this emergent space, so “AI agent” might mean different things to different people. Infrastructure is also still being built out to deliver on its envisaged capabilities. But the basic concept implies an autonomous system that may draw on multiple AI systems to carry out multistep tasks. Think of API endpoints as “buttons” on the back of a piece of software that other programs can press to make it do things. Developers use these interfaces to build integrations — for example, allowing one application to pull data from another, or enabling an AI agent to control third-party services directly without a human manually operating each interface. Most smart home devices and connected platforms have these hidden buttons available, even if ordinary users never see or interact with them. As AI agents grow more capable, they are increasingly able to find and use these endpoints on their own, opening up powerful — and sometimes unexpected — possibilities for automation. Given a simple question, a human brain can answer without even thinking too much about it — things like “which animal is taller, a giraffe or a cat?” But in many cases, you often need a pen and paper to come up with the right answer because there are intermediary steps. For instance, if a farmer has chickens and cows, and together they have 40 heads and 120 legs, you might need to write down a simple equation to come up with the answer (20 chickens and 20 cows). In an AI context, chain-of-thought reasoning for large language models means breaking down a problem into smaller, intermediate steps to improve the quality of the end result. It usually takes longer to get an answer, but the answer is more likely to be correct, especially in a logic or coding context. Reasoning models are developed from traditional large language models and optimized for chain-of-thought thinking thanks to reinforcement learning. (See:Large language model) This is a more specific concept that an “AI agent,” which means a program that can take actions on its own, step by step, to complete a goal. A coding agent is a specialized version applied to software development. Rather than simply suggesting code for a human to review and paste in, a coding agent can write, test, and debug code autonomously, handling the kind of iterative, trial-and-error work that typically consumes a developer’s day. These agents can operate across entire codebases, spotting bugs, running tests, and pushing fixes with minimal human oversight. Think of it like hiring a very fast intern who never sleeps and never loses focus — though, as with any intern, a human still needs to review the work. Although somewhat of a multivalent term, compute generally refers to the vitalcomputational powerthat allows AI models to operate. This type of processing fuels the AI industry, giving it the ability to train and deploy its powerful models. The term is often a shorthand for the kinds of hardware that provides the computational power — things like GPUs, CPUs, TPUs, and other forms of infrastructure that form the bedrock of the modern AI industry. A subset of self-improving machine learning in which AI algorithms are designed with a multi-layered, artificial neural network (ANN) structure. This allows them to make more complex correlations compared to simpler machine learning-based systems, such as linear models or decision trees. The structure of deep learning algorithms draws inspiration from the interconnected pathways of neurons in the human brain. Deep learning AI models are able to identify important characteristics in data themselves, rather than requiring human engineers to define these features. The structure also supports algorithms that can learn from errors and, through a process of repetition and adjustment, improve their own outputs. However, deep learning systems require a lot of data points to yield good results (millions or more). They also typically take longer to train compared to simpler machine learning algorithms — so development costs tend to be higher. (See:Neural network) Diffusion is the tech at the heart of many art-, music-, and text-generating AI models. Inspired by physics,diffusion systems slowly “destroy” the structure of data— for example, photos, songs, and so on — by adding noise until there’s nothing left. In physics, diffusion is spontaneous and irreversible — sugar diffused in coffee can’t be restored to cube form. But diffusion systems in AI aim to learn a sort of “reverse diffusion” process to restore the destroyed data, gaining the ability to recover the data from noise. Distillation is a technique used to extract knowledge from a large AI model with a ‘teacher-student’ model. Developers send requests to a teacher model and record the outputs. Answers are sometimes compared with a dataset to see how accurate they are. These outputs are then used to train the student model, which is trained to approximate the teacher’s behavior. Distillation can be used to create a smaller, more efficient model based on a larger model with a minimal distillation loss. This is likely how OpenAI developed GPT-4 Turbo, a faster version of GPT-4. While all AI companies use distillation internally, it may have also been used by some AI companies to catch up with frontier models. Distillation from a competitor usuallyviolatesthe terms of service of AI API and chat assistants. This refers to the further training of an AI model to optimize performance for a more specific task or area than was previously a focal point of its training — typically by feeding in new, specialized (i.e., task-oriented) data. Many AI startups are taking large language models as a starting point to build a commercial product but are vying to amp up utility for a target sector or task by supplementing earlier training cycles with fine-tuning based on their own domain-specific knowledge and expertise. (See:Large language model [LLM]) A GAN, or Generative Adversarial Network, is a type of machine learning framework that underpins some important developments in generative AI when it comes to producing realistic data — including (but not only) deepfake tools. GANs involve the use of a pair of neural networks, one of which draws on its training data to generate an output that is passed to the other model to evaluate. The two models are essentially programmed to try to outdo each other. The generator is trying to get its output past the discriminator, while the discriminator is working to spot artificially generated data. This structured contest can optimize AI outputs to be more realistic without the need for additional human intervention. Though GANs work best for narrower applications (such as producing realistic photos or videos), rather than general purpose AI. Hallucination is the AI industry’s preferred term for AI models making stuff up — literally generating information that is incorrect. Obviously, it’s a huge problem for AI quality. Hallucinations produce GenAI outputs that can be misleading and could even lead to real-life risks — with potentially dangerous consequences (think of a health query that returns harmful medical advice). The problem of AIs fabricating information is thought to arise as a consequence of gaps in training data. Hallucinations are contributing to a push toward increasingly specialized and/or vertical AI models — i.e. domain-specific AIs that require narrower expertise — as a way to reduce the likelihood of knowledge gaps and shrink disinformation risks. Inference is the process of running an AI model. It’s setting a model loose to make predictions or draw conclusions from previously seen data. To be clear, inference can’t happen without training; a model must learn patterns in a set of data before it can effectively extrapolate from this training data. Many types of hardware can perform inference, ranging from smartphone processors to beefy GPUs to custom-designed AI accelerators. But not all of them can run models equally well. Very large models would take ages to make predictions on, say, a laptop versus a cloud server with high-end AI chips. [See:Training] Large language models, or LLMs, are the AI models used by popular AI assistants, such asChatGPT,Claude,Google’s Gemini,Meta’s AI Llama,Microsoft Copilot, orMistral’s Le Chat. When you chat with an AI assistant, you interact with a large language model that processes your request directly or with the help of different available tools, such as web browsing or code interpreters. LLMs are deep neural networks made of billions of numerical parameters (or weights, see below) that learn the relationships between words and phrases and create a representation of language, a sort of multidimensional map of words. These models are created from encoding the patterns they find in billions of books, articles, and transcripts. When you prompt an LLM, the model generates the most likely pattern that fits the prompt. (See:Neural network) Memory cache refers to an important process that boosts inference (which is the process by which AI works to generate a response to a user’s query). In essence, caching is an optimization technique, designed to make inference more efficient. AI is obviously driven by high-octane mathematical calculations and every time those calculations are made, they use up more power. Caching is designed to cut down on the number of calculations a model might have to run by saving particular calculations for future user queries and operations. There are different kinds of memory caching, although one of the more well-known isKV (or key value) caching. KV caching works in transformer-based models, and increases efficiency, driving faster results by reducing the amount of time (and algorithmic labor) it takes to generate answers to user questions. (See:Inference) Model Context Protocol, or MCP, is an open standard that lets AI models connect to outside tools and data — your files, databases, or apps like Slack and Google Drive — without a developer building a custom connector for every single pairing. Think of it as a USB-C port for AI. Anthropic introduced MCP in 2024 and later handed it over to the Linux Foundation, and it’s since been adopted by OpenAI, Google, and Microsoft, making it one of the fastest-spreading standards in recent AI history. Mixture of Experts is a model architecture that splits a neural network into many smaller specialized sub-networks, or “experts,” and only activates a handful of them for any given task. Rather than routing every request through the entire model — like calling in your whole office for every question — an MoE model has a built-in “router” that picks just the right specialists for the job. This makes it possible to build enormous models that stay relatively fast and cheap to run, since only a fraction of the network is doing work at any one time. Mistral AI’s Mixtral model is a well-known example; OpenAI’s newer GPT models are also widely believed to use some version of this approach, though the company has never officially confirmed it. (See:Neural network,Deep learning) A neural network refers to the multi-layered algorithmic structure that underpins deep learning — and, more broadly, the whole boom in generative AI tools following the emergence of large language models. Although the idea of taking inspiration from the densely interconnected pathways of the human brain as a design structure for data processing algorithms dates all the way back to the 1940s, it was the much more recent rise of graphical processing hardware (GPUs) — via the video game industry — that really unlocked the power of this theory. These chips proved well suited to training algorithms with many more layers than was possible in earlier epochs — enabling neural network-based AI systems to achieve far better performance across many domains, including voice recognition, autonomous navigation, and drug discovery. (See:Large language model [LLM]) Open source refers to software — or, increasingly, AI models — where the underlying code is made publicly available for anyone to use, inspect, or modify. In the AI world, Meta’s Llama family of models is a prominent example; Linux is the famous historical parallel in operating systems. Open source approaches allow researchers, developers, and companies around the world to build on top of one another’s work, accelerating progress and enabling independent safety audits that closed systems cannot easily provide. Closed source means the code is private — you can use the product but not see how it works, as is the case with OpenAI’s GPT models — a distinction that has become one of the defining debates in the AI industry. Parallelization means doing many things at the same time instead of one after another — like having 10 employees working on different parts of a project at the same time instead of one employee doing everything sequentially. In AI, parallelization is fundamental to both training and inference: modern GPUs are specifically designed to perform thousands of calculations in parallel, which is a big reason why they became the hardware backbone of the industry. As AI systems grow more complex and models grow larger, the ability to parallelize work across many chips and many machines has become one of the most important factors in determining how quickly and cost-effectively models can be built and deployed. Research into better parallelization strategies is now a field of study in its own right. RAMageddon is the fun new term for a not-so-fun trend that is sweeping the tech industry: an ever-increasing shortage of random access memory, or RAM chips, which power pretty much all the tech products we use in our daily lives. As the AI industry has blossomed, the biggest tech companies and AI labs — all vying to have the most powerful and efficient AI — are buying so much RAM to power their data centers that there’s not much left for the rest of us. And that supply bottleneck means that what’s left is getting more and more expensive. That includes industries like gaming (where major companies have had toraise prices on consolesbecause it’s harder to find memory chips for their devices), consumer electronics (where memory shortage could causethe biggest dip in smartphone shipmentsin more than a decade), and general enterprise computing (because those companies can’t get enough RAM for their own data centers). The surge in prices is only expected to stop after the dreaded shortage ends but, unfortunately, there’snot really much of a signthat’s going to happen anytime soon. Like AGI, recursive self-improvement is a threshhold for how smart AI can get, and how little it may rely on humans. In the RSI scenario, AI models start improving themselves without human intervention, leading to a huge acceleration in capabilities and autonomy. In some tellings, this would be a cataclysmic moment akin to the singularity, a moment when AI models become immune to outside intervention. But RSI also describes a basic capability — can an AI model design its own successor? — which makes it much easier for engineers to try to build it.A number of recent AI startupshave set out to build recursively self-improving models, but most of them dismiss the apocalyptic implications, presenting RSI as simply the next frontier for research. Reinforcement learning is a way of training AI where a system learns by trying things and receiving rewards for correct answers — like training your beloved pet with treats, except the “pet” in this scenario is a neural network and the “treat” is a mathematical signal indicating success. Unlike supervised learning, where a model is trained on a fixed dataset of labeled examples, reinforcement learning lets a model explore its environment, take actions, and continuously update its behavior based on the feedback it receives. This approach has proven especially powerful for training AI to play games, control robots, and, more recently, sharpen the reasoning ability of large language models. Techniques like reinforcement learning from human feedback, or RLHF, are now central to how leading AI labs fine-tune their models to be more helpful, accurate, and safe. When it comes to human-machine communication, there are some obvious challenges — people communicate using human language, while AI programs execute tasks through complex algorithmic processes informed by data. Tokens bridge that gap: they are the basic building blocks of human-AI communication, representing discrete segments of data that have been processed or produced by an LLM. They are created through a process called tokenization, which breaks down raw text into bite-sized units a language model can digest, similar to how a compiler translates human language into binary code a computer can understand. In enterprise settings, tokens also determine cost — most AI companies charge for LLM usage on a per-token basis, meaning the more a business uses, the more it pays. So again, tokens are the small chunks of text — often parts of words rather than whole ones — that AI language models break language into before processing it; they are roughly analogous to “words” for the purposes of understanding AI workloads. Throughput refers to how much can be processed in a given period of time, so token throughput is essentially a measure of how much AI work a system can handle at once. High token throughput is a key goal for AI infrastructure teams, since it determines how many users a model can serve simultaneously and how quickly each of them receives a response. AI researcher Andrej Karpathy has described feeling anxious when his AI subscriptions sit idle — echoing the feeling he had as a grad student when expensive computer hardware wasn’t being fully utilized — a sentiment that captures why maximizing token throughput has become something of an obsession in the field. Developing machine learning AIs involves a process known as training. In simple terms, this refers to data being fed in in order that the model can learn from patterns and generate useful outputs. Essentially, it’s the process of the system responding to characteristics in the data that enables it to adapt outputs toward a sought-for goal — whether that’s identifying images of cats or producing a haiku on demand. Training can be expensive because it requireslotsof inputs, and the volumes required have been trending upwards — which is why hybrid approaches, such as fine-tuning a rules-based AI with targeted data, can help manage costs without starting entirely from scratch. [See:Inference] A technique where a previously trained AI model is used as the starting point for developing a new model for a different but typically related task — allowing knowledge gained in previous training cycles to be reapplied. Transfer learning can drive efficiency savings by shortcutting model development. It can also be useful when data for the task that the model is being developed for is somewhat limited. But it’s important to note that the approach has limitations. Models that rely on transfer learning to gain generalized capabilities will likely require training on additional data in order to perform well in their domain of focus (See:Fine tuning) Validation loss is a number that tells you how well an AI model is learning during training — and lower is better. Researchers track it closely as a kind of real-time report card, using it to decide when to stop training, when to adjust hyperparameters, or whether to investigate a potential problem. One of the key concerns it helps flag is overfitting, a condition in which a model memorizes its training data rather than truly learning patterns it can generalize to new situations. Think of it as the difference between a student who genuinely understands the material and one who simply memorized last year’s exam — validation loss helps reveal which one your model is becoming. Weights are core to AI training, as they determine how much importance (or weight) is given to different features (or input variables) in the data used for training the system — thereby shaping the AI model’s output. Put another way, weights are numerical parameters that define what’s most salient in a dataset for the given training task. They achieve their function by applying multiplication to inputs. Model training typically begins with weights that are randomly assigned, but as the process unfolds, the weights adjust as the model seeks to arrive at an output that more closely matches the target. For example, an AI model for predicting housing prices that’s trained on historical real estate data for a target location could include weights for features such as the number of bedrooms and bathrooms, whether a property is detached or semi-detached, whether it has parking, a garage, and so on. Ultimately, the weights the model attaches to each of these inputs reflect how much they influence the value of a property, based on the given dataset. This article is updated regularly with new information.
View
The browser wars aren’t about search anymore — here are the best alternatives to Chrome and Safari
The browser wars have entered a new phase this year: the fight isn’t just over search results anymore, it’s over which company’s AI gets to act on your behalf inside the browser itself. Google Chrome and Apple’s Safari still dominate the market overall, with Chrome’s edge coming largely from how aggressively it has woven generative AI into search. But 2026 has brought a wave of new entrants — from well-funded startups to Big Tech itself — all betting that the browser is about to become less like a window onto the web and more like an assistant that gets things done for you. Users looking for alternatives to Chrome and Safari can choose from a growing variety of browsers aimed at challenging the industry giants. To help navigate the competitive landscape, we’ve compiled an overview of some of the top alternative browsers available today. This includes browsers leveraging AI, open source browsers that promote customization and privacy, and “mindful browsers” — a new term that refers to browsers designed to enhance user well-being. Perplexity is the most recent startup in the space tolaunch an AI-powered web browser. CalledComet, the company’s new product acts as a chatbot-based search engine, and can perform actions like summarizing emails, browsing web pages, and performing tasks such as sending calendar invites. It’s currently only available to users with Perplexity’s $200/month Max plan, but there’s also a waitlist where people can sign up. The Browser Company, the startup behind the Arc browser,recently introducedDia, its AI-centric browser that looks similar to Google Chrome but with an AI chat tool. Currently available as aninvite-only beta, Dia is designed to help users navigate the web more easily. It’s able to look at every website that a user has visited and every website they’re logged into, enabling it to help you find information and perform tasks. For instance, Dia can provide information about the page a user is currently browsing, answer questions about a product, and summarize uploaded files. To get early access to Dia, users have to be an Arc member. Non-members can join the waitlist. Anotherrecent entryinto the AI agentic browser war is Opera’sNeon, which has contextual awareness and can do things like researching, shopping, and writing snippets of code. Notably, it can even perform tasks while the user is offline. Neon is currently available on macOS and Windows. The subscription costs $19.90 per month. OpenAI recently launched its AI-powered web browser, calledAtlas. The browser allows users to ask ChatGPT about search results and browse websites within the chatbot instead of being directed to outside links. There’s also an “agent mode” for users to ask ChatGPT to complete tasks on their behalf. Atlas was first rumored to launch inJuly; however, it only became available on macOS in October. It’s expected to arrive on Windows, iOS, and Android devices soon. Backed by Y Combinator,Asideis an upcoming AI-first, browser-native automation platform built to autonomously complete tasks, fill out forms, and manage data on behalf of users. The company describes the experience simply: “Give it your passwords, browsing history, and browser context.” Unlike traditional automation tools that rely on integrations, Aside operates directly within the browser itself, allowing it to work across Gmail, Notion, Slack, Figma, and banking platforms. Users can sign up for the waitlist ahead of launch. Jatterlaunched its AI-powered browser in June, giving users the ability to ask questions about any webpage, uncover relevant insights, and receive personalized recommendations based on their browsing activity. Additionally, Jatter offers an integrated Notes app, so it can learn from that content, summarize notes, and surface key details. Jatter is currently available on Mac, Windows, iOS, and Android devices. It’s free to use, but there’s also an optional subscription for $10 per month. Braveis among the more well-knownprivacy-first browsers, popular for its built-in ad and tracker blocking capabilities. It also has a gamified approach to browsing, rewarding users with its own cryptocurrency called Basic Attention Token (BAT). When users choose to opt in to view ads, supporting their favorite websites, they get a share of the ad revenue. Additional features include a VPN service,an AI assistant, anda video calling feature. DuckDuckGois anotherbrowserthat many people are probably already familiar with, thanks to its search engine by the same name. Launched in 2008, the company recently made significant investments in its browser to stay competitive byintroducing generative AI features, such as a chatbot. It alsoenhanced its scam blockerto detect a wider range of scams, including fake cryptocurrency exchanges, scareware tactics, and fraudulent e-commerce websites. In addition to blocking scams, DuckDuckGo prevents trackers and ads, and it doesn’t track user data, resulting in fewer pop-ups for users. Ladybird, led by GitHub co-founder and former CEO Chris Wanstrath, has an ambitious mission compared to other rivals: It aims to build an entirely new open source browser from scratch. This means it will not rely on code from existing browsers, a feat that has rarely been accomplished. Most alternative web browsers depend on the Chromium open source project maintained by Google, which is the most widely used base for many browsers. Like other privacy-focused browsers, Ladybird will offer features to minimize data collection, such as a built-in ad blocker and the ability to block third-party cookies. The browser has yet to be launched, with an alpha version scheduled for release in 2026 for early adopters, available on Linux and macOS. Vivaldiis a Chromium-basedbrowsercreated by one of the original developers of the Opera browser. Its biggest selling point is its customizable user interface, which allows users to change the appearance and enable or disable features. One unique feature is that the browser window changes color to match the website being viewed. Other key features include ad blocking, a password manager, no user data tracking, and productivity tools such as a calendar and notes. Operalaunchedthe Air browser in February, becoming one of the first mindfulness-themed browsers in the space. WhileOpera Airfunctions like a typical web browser, it includes unique features designed to support mental well-being. These features consist of break reminders and breathing exercises. Another feature, called “Boosts,” provides a selection of binaural beats to either help improve focus or relaxation. SigmaOSis a Mac-only browser featuring a workspace-style interface that emphasizes productivity. It displays tabs vertically, allowing users to treat them like a to-do list that can be marked as complete or snoozed for later. Users can create workspaces — essentially groups of tabs — to better organize different activities, such as separating work from entertainment. This Y Combinator-backed browser hasbeen aroundfor a few years now and has most recently begun introducing moreAI features,including the ability to summarize various elements of a web page, such as ratings, reviews, and prices. It also has anAI assistantthat can answer questions, translate text, and rewrite content. SigmaOS is free to use, but users who want more than three workspaces can subscribe to a plan for $8 per month, which provides unlimited workspaces. Zen Browseraims to create a “calmer internet” with its open source browser. Zen lets users organize tabs into Workspaces, and offers Split View to view two tabs side by side, among other productivity-focused features. Users can also enhance their browsing experience with community-made plug-ins and themes, such as a mod that makes the tab background transparent. This story has been updated after publication to include newly launched browsers.
View
Why AI Startups Are Knocking on Corporate Doors Instead of VC Offices
MNCs are redesigning their accelerator programmes around business collaboration rather than investment.
View
NIIT Brings StackRoute, RPS Consulting Under New Enterprise Tech Business
The move aims to address changing enterprise talent needs by helping organisations build future-ready technology talent and manage large-scale workforce transformation.
View