Category: Top News

IBL News | New York

OpenClaw released its largest update yesterday, comprising over 16,000 pull requests and built by 933 contributors, including 569 first-time contributors.

This update, which took nearly two months, touches every part of OpenClaw, including installation, messaging, memory, skills, models, automations, the browser and native apps, plugins, security, and a very long tail of fixes.

“We started by simplifying installation and rebuilding the browser app as a first-class experience, but doing that properly meant carrying the cleanup through the rest of OpenClaw until it became OpenClaw 2.0,” said the organization behind OpenClaw.

OpenClaw 2.0 installs starts with what is already on someone’s computer, including existing ChatGPT or Claude subscriptions, API keys, and local models.

“We cut or simplified a lot of configuration and moved the rest out of initial setup, letting people get to a first conversation faster and finish setting up their Claw by talking to it.”

IBL News | New York

Salesforce will embed its entire CRM within Claude via a plugin called Claudeforce, with an open beta planned for September. On Tuesday, the San Francisco-based SaaS company announced a partnership with Anthropic’s Claude.

The plugin, within Claude CoWork, ships with 37 pre-built sales skills and will allow sellers to query, update, and act on live CRM data without opening Salesforce.

“We’re launching Claudeforce, running directly on top of Salesforce via our new AIforce UI harness, Headless 360, Data 360, Tableau, and Slack. Here, the UI is the AI — allowing you to build custom apps dynamically and answer any enterprise question,” Marc Benioff, Salesforce’s Chair and CEO, said in a statement.

“By fusing Claude’s extraordinary reasoning with the trusted data, workflows, and governance every enterprise runs on, we’re delivering a dynamic interface that thinks, reasons, and acts. This is how every business will run.”

The road to Claudeforce began in March 2026. At its TDX developer conference, Salesforce released Headless 360, a collection of APIs, MCP servers, and command-line tools that let AI agents call Salesforce data, workflows, and governance rules directly, without a user interface.

Salesforce believes that its value is not the application or the UI, but the data and metadata, the years’ worth of encoded workflows and business practices built up inside Salesforce.

With Claude CoWork, Salesforce’s productivity increases dramatically. A seller’s morning ritual — deciding which opportunities to work — traditionally involves opening an opportunities list, clicking into each record, reading activities and meeting histories, and mentally synthesizing everything, a process that requires many clicks inside Salesforce. Claude will execute all of that.

The announcement raises questions about Agentforce, the agent platform Salesforce has spent roughly two years promoting as its AI centerpiece.

The company now says that Agentforce is a kind of help.salesforce.com, designed to interface with the end customer, while Claudeforce  — Salesforce in Claude — is a knowledge worker agent, built for salespeople, to help them do their day-to-day job without having to do the traditional part of their job, which is clicking around in user interfaces and trying to synthesize data themselves.

IBL News | New York

Nvidia has agreed to buy Hugging Face for $12.9 billion, The Information reported. The companies avoided commenting on the transaction.

Hugging Face, founded in 2016, is one of the most popular hubs where developers share and download open-source AI models.

Buying it will give Nvidia a strong presence in the open-source AI space, helping to protect its dominance in AI chips, especially since the biggest closed-source companies (OpenAI, Google, Amazon, and Anthropic) are now building their own AI chips to reduce their reliance on Nvidia.

Also, owning Hugging Face — which already helps developers run their AI models using rented computing power — could give Nvidia a way back into the cloud computing market without starting from scratch, since Nvidia reportedly scaled back its own cloud business, called DGX Cloud, about a year ago.

On its side, Nvidia has already poured tens of billions of dollars into building its own open-source AI models.

Not surprisingly, Hugging Face CEO Clem Delangue has spent much of this year publicly aligned with Nvidia’s open-source push. In this regard, Hugging Face used an Nvidia-modified version of a Chinese open-source model to defend itself after a cyberattack.

On the other hand, Chinese labs like Moonshot AI had released systems — such as its Kimi K3 model — that matched leading U.S. models on benchmarks while costing much less to run.

The price marks a huge jump from Hugging Face’s last known value. The company raised $235 million in a 2023 funding round, valuing it at $4.5 billion. That round was led by Salesforce Ventures, with money also coming from Alphabet’s GV, IBM Ventures, and Nvidia, among others.

Hugging Face generates only about $150 million in annual revenue, up from roughly $100 million just two months earlier.

Nvidia acquires Hugging Face for ~$12.9bn. But who gets the 💰? My usual breakdown below 👇

*Investors are sharing ~$7.3bn of profits on <$400m invested. A cool ~20x blended.*

*1) The single biggest winner here? Lux Capital.* 👑

Lux is taking the crown for the most $ returned… pic.twitter.com/AScos7vSee

— Paul (@P_Bonnet) August 27, 2026

IBL News | New York

Perplexity announced a version of its agentic “Computer platform”, named Portable Computer, that runs on hardware owned by clients, such as NVIDIA’s DGX Spark and Linux-powered systems with RTX GPUs.

The launch, developed in close partnership with Nvidia, reflects a new trend of AI agent workloads moving off the cloud and onto local devices with no billing credit consumption.

This local app has the same UI as Perplexity’s cloud version.

Portable Computer will be offered for Pro, Max, Enterprise Pro, and Enterprise Max subscribers on Linux, with Windows support following in September. Any RTX GPU with at least 24GB of VRAM — roughly a GeForce RTX 3090 or newer — clears the bar.

While Perplexity Computer orchestrates AI models, files, tools, and web access to complete multi-step tasks — reviewing folders of documents, analyzing data, producing reports, and pushing results into business systems. Portable Computer replicates that experience locally: the local models, agent harness, inference engine, tools, app connectors, and a security sandbox come packaged together in a single system.

Typically, today, with local AI stacks users must assemble and operate those pieces separately — downloading model weights, standing up an inference server, wiring together tools, and tuning performance.

Venture Beat did a demo running a 27-billion-parameter Qwen model at full GPU utilization on a DGX Spark, the agent reviewed each document and flagged cases where the hypothetical investor was paying unnecessary fees.

Alongside the launch, Perplexity published a research paper arguing that effective local agents require the model and the agent harness — the scaffolding of prompts, tools, and orchestration logic around the model — to be designed together. The core insight: general-purpose harnesses assume a frontier model that can absorb enormous contexts, navigate sprawling tool surfaces, and plan over long horizons. Small local models buckle under those demands.

Perplexity found empirically that although models like Qwen 3.8 27B advertise 260,000-token context windows, they begin to struggle beyond 100,000 tokens. So the company built a deliberately minimal harness: a succinct system prompt, a small set of core tools, and capabilities that load and unload as on-demand “skills” rather than sitting permanently in context.

IBL News | New York

Developers have published over 100,000 variants of Google’s open-source Gemma model over the past two years, including offline educational hubs for disconnected regions. This software has registered over 1 billion downloads.

Building on this momentum, Google has introduced the “Awesome Gemma” GitHub repository, a curated directory featuring the best community projects, fine-tunes, tutorials, and developer tools.

“Developers have proven Gemma can deliver complex reasoning in some of the most constrained and extreme environments imaginable, building a thriving ecosystem of innovation we call the Gemmaverse,” Google said in a blog post.

• In India, the National Health Authority (NHA) integrated Gemma 4 and the Google open-source Medical Data Toolkit into Aarogya Setu 2.0, an app with over 100 million downloads on Android.

By using Gemma 4 to process complex medical reports into standardized digital formats, the app helps citizens manage and securely share their health data across providers. This capability demonstrates how developers can deploy Google’s open models to handle critical information at a massive scale.

• Researchers from Yale and Google built C2S-Scale, a powerful AI model designed to interpret the “language” of single cells. Built on Gemma, C2S-Scale successfully discovered a novel cancer therapy pathway that was verified in living cells. This research marks the first time an AI system produced novel mechanistic therapeutic pathways that were verified in living cells.

• Google’s medical model, MedGemma, is helping developers accelerate the development of their healthcare applications, from improving workflow management and patient communication to supporting diagnostics and treatment. Today, MedGemma is being used by organizations worldwide to develop real-world clinical applications, from supporting outpatient triage at the All India Institute of Medical Sciences (AIIMS) to creating systems that support frontline health workers in rural Uganda.

• In collaboration with Georgia Tech and the Wild Dolphin Project, researchers developed DolphinGemma. This specialized AI model processes complex dolphin vocalizations to predict sequences of sounds. Researchers are actively iterating on this complex challenge, and it’s just the beginning. So stay tuned for much more to come from under the sea.

• Gemma Challenge on Kaggle, where the community submitted over 1600 projects aimed at solving real-world problems.

IBL News | New York

Meta is planning to launch its version of OpenClaw, named “Hatch”, in late August or early September, and its latest AI model, Watermelon, in October.

These are the two models of Mark Zuckerberg’s push for “personal superintelligence.” In an August 10th essay, he said Meta would build personal agents and models for billions of people, offer free versions, and charge users who want more computing capacity.

In this regard, Hatch will be a consumer AI agent inspired by open-source OpenClaw, according to a report from The Information.

Hatch has been trained to access services including DoorDash, Etsy, Reddit, Yelp, and Outlook. Early prototypes include a customizable dashboard where users can see tools generated by the agent, such as a fitness tracker or travel itinerary.

OpenClaw runs on the user’s own machine and supports add-on capabilities called skills, but installing and configuring the system still requires command-line work and decisions about models, credentials, and permissions.

Hatch subscription might cost 199.99 per month, and it seems that the company will target professionals and heavy users willing to pay for an agent that can build small software tools and complete long-running tasks across multiple services.

On August 5th, Meta released Muse Spark 1.2 and the Muse Code terminal agent. Five days later, Meta published Muse Glimmer, a 30-billion-parameter open-weight model designed for local agent workflows. Meta already has distribution through WhatsApp, Instagram, Facebook, and its standalone AI app.

IBL News | New York

Anthropic’s revenue continues to grow at an accelerated rate, surpassing $65 billion at the end of July, up from $47 billion in May and just $9 billion at the end of last year. 

Investors expect Anthropic to continue growing at approximately the same rate for the remainder of the year, finishing 2026 between $100 billion and $120 billion, according to Bloomberg. 

After completing a $65 billion round, Anthropic was last valued at $965 billion in late May. The San Francisco-based company is expected to go public this fall, ahead of OpenAI. It will seek a public valuation of $2 trillion or more, which would make it the largest market debut on record.

Meanwhile, rival OpenAI has doubled its revenue to $40 billion, up from $20 billion at the end of 2025, Bloomberg reported last week.

The two companies calculate their revenue metrics differently, but Anthropic’s growth rate has captivated investors far more than OpenAI’s has.

IBL News | New York

AI agents are undermining online education by taking courses on learners’ behalf, quickly accelerating cheating to earn certificates.

At many community colleges and globally renowned universities, classes are self-paced with prerecorded lectures, also known as asynchronous. Assignments and tests are submitted digitally.

In fact, half of American college students took at least one class last year, up from about a third in 2019. Sometimes students can earn full degrees online. In other cases, they take individual classes.

To prevent misuse of an unregulated technology, many educators are shifting to oral presentations and handwritten exams in physical classrooms.

Beyond the copy-and-pasting of chatbot-produced essays and problem sets, AI agents can watch lecture videos, take tests, write papers, and even chat with classmates about reading assignments.

This is happening in a context where education leaders are eager to embrace a technology that could reshape the world of work.

In addition, AI detection software can make mistakes, so there is no concrete proof of misconduct. And some administrators, professors fear, do not want to risk losing a paying customer in an era of tightening budgets and shifting demographics.

In tests by The New York Times, none of the three leading AI tools among students — ChatGPT, Gemini, and Grammarly — refused to write a paper on a student’s behalf. Also, AI companies have not prevented agents from logging into learning platforms such as Canvas, Brightspace, and Blackboard, which colleges use to manage online classes.

Experts have warned that more novel forms of AI cheating are emerging. Wearable devices, like glasses, can allow students to request help even while working in a secure web browser. AI avatars can impersonate students over video.

Some institutions are also changing how online assignments are designed, leaning on personal responses that are more difficult for AI to fake.

• The New York Times: AI Agents Are Taking Entire Online Courses for Cheating Students

IBL News | New York

OpenRouter, which provides developers with a single gateway to more than 400 AI models, agreed to be acquired by Stripe Inc. for more than $7 billion, according to a Bloomberg report — although Stripe hasn’t confirmed it.

OpenRouter’s success is based on switching between OpenAI, Anthropic, or a cheaper open-weight alternative without touching the code, with the router choosing based on cost, speed, or which provider is up. It takes about 5% of the inference spend passing through it and says it has roughly 8 million users.

OpenRouter was founded in 2023 by Alex Atallah and Louis Vichy. Atallah described the company as “Stripe for AI,” arguing that a single access point across model providers removes integration work and blocks lock-in.

Weekly throughput hit 25 trillion tokens by May, five times the figure six months earlier. Revenue is about $50 million annualized as of March, against roughly $19 million when 2025 closed.

Alphabet Inc.’s growth fund, CapitalG LP, led the $113 million Series B that OpenRouter announced on May 26, with Andreessen Horowitz and Menlo Ventures alongside. The round was raised on a $1.3 billion valuation, more than double the figure a year earlier.

Stripe co-authored the Agentic Commerce Protocol with OpenAI and has been OpenRouter’s payments provider since at least January, when the two announced a token-billing integration that meters model usage and automatically prices it.

OpenRouter competes with open-source routing projects such as LiteLLM and with the routing features set by major cloud providers.

Stripe paid about $1.1 billion for stablecoin infrastructure startup Bridge Network Inc. in February 2025 and also acquired crypto wallet provider Privy Inc. In July, it joined private equity firm Advent International L.P. in a bid of more than $53 billion to take PayPal Holdings Inc. private. A February tender offer valued Stripe at $159 billion.

IBL News | New York

Crypto firm World Liberty Financial, which is 38% owned by the Trump family, is collaborating with WorldClaw venture, a Hong Kong-based company, by offering Chinese AI models, Reuters reported. The collaboration, which is not illegal, seems to be based mainly on the fact that Chinese WorldClaw allows customers to accept World Liberty’s crypto tokens as payment.

The White House spokesperson Anna Kelly said in a statement that “there are no conflicts of interest” in the relationship between World Liberty and WorldClaw and that “President Trump ​only acts in the best interests of the American public.”

David Wachsman, a spokesman for World Liberty Financial, said WorldClaw was an independent company and noted that major U.S. firms offer AI from both Chinese and American tech companies. “This is a common and widely accepted approach,” he said in an email.

According to Reuters, 43 of WorldClaw’s 90 AI models have been flagged by U.S. administrations for national security concerns.

Behind these models, which are typically less expensive and have increasing global traction, are Alibaba, Baidu, Z.ai, and other Chinese technology companies. However, WorldClaw ‌also offers access to dozens of models from U.S. firms such as OpenAI and Anthropic.

Seven experts on Chinese technology, trade, and government ethics, quoted by Reuters, said the business collaboration runs counter to the Trump administration’s stance on Chinese tech companies.