Рубрика: AI

TCS has rolled out a new AI-powered upgrade to its TCS ADD Risk-Based Quality Management (RBQM) platform, designed to give drugmakers and research organisations sharper, real-time oversight of clinical trials. The platform aims to help teams spot risks earlier, improve data quality, and manage increasingly complex trial setups.

The new version introduces four AI-driven modules focused on risk assessment, quality tolerance limits, trial analytics, and subject-level data monitoring. Together, these tools allow researchers to detect problems much sooner than traditional monitoring methods.

TCS says these modules are among the few globally that are fully interoperable and can be customised to suit different trial designs. This feature can shorten deployment time for sponsors.

“In today’s rapidly evolving clinical research environment, traditional approaches to quality management are no longer sufficient,” said Rachna Malik, Global Head of TCS ADD. She said the upgraded platform supports faster, data-backed decisions and can help bring new therapies to patients more quickly.

Industry trends support TCS’s push toward AI-driven oversight. Adoption of AI in India’s healthcare GCCs has risen sharply from 65% in 2019 to 86% in 2024, Zinnov managing partner Karthik Padmanabhan told AIM, noting that AI tools are now central to improving patient recruitment, monitoring risks, and ensuring regulatory compliance.

The update comes as the life sciences industry increasingly relies on AI and analytics to navigate stricter regulations and the challenges of decentralised, adaptive trials. TCS notes that the platform is aligned with international guidelines ICH E6(R2) and the upcoming E6(R3), and incorporates Quality by Design principles from the start of a study through execution.

TCS says the platform has been used in more than 1,300 studies across 32,000 sites, a sign that AI-driven oversight is fast becoming a standard part of modern clinical research.

The post TCS Launches AI Platform to Oversee Clinical Trial Oversight appeared first on Analytics India Magazine.

Sentient AI’s launch of Recursive Open Meta-Agent (ROMA), in September 2025 marks a pivotal moment in the evolution of open-source AI frameworks in India. Far from being “just another model,” ROMA represents a new way of thinking about how AI systems should reason, coordinate, and solve complex problems.

Built as a hierarchical meta-agent framework, ROMA abandons the monolithic design of traditional large models and instead orchestrates a structured tree of specialised agents, each responsible for solving part of a larger task.

The result? A transparent, auditable, high-performance system that challenges the dominance of closed AI architectures and sets a new direction for the future of decentralised intelligence.

What ROMA Actually Is, and Why It Matters

As Himanshu Tyagi, co-founder of Sentient, told AIM, “ROMA isn’t a model; it’s a reasoning architecture. That distinction is crucial as models generate outputs, while reasoning architectures determine how those models think, plan, and coordinate to solve problems.”

ROMA’s defining innovation lies in its recursive hierarchical task tree. At its foundation is an “Atomiser” that decides whether a task is atomic, directly executable, or requires further planning. If planning is required, control moves to the “Planner,” which decomposes the goal into subtasks, each of which is recursively fed back into the Atomiser. This allows ROMA to scale to long-horizon, multi-step reasoning tasks that typically confound single-model systems.

This architecture is not only algorithmically efficient but also involves minimal complexity structurally. As Tyagi said, “The algorithmic innovation is elegant yet simple (simple recursion), leading to phenomenal performance improvements.”

By treating reasoning as a planning problem rather than a generation problem, ROMA reduces compounding errors, improves long-term coherence, and enables parallel execution, allowing independent subtasks to be solved simultaneously, the founders said.

Tyagi argued that the narrative around openness and safety, often used by Big Tech to justify closed systems, is fundamentally flawed. “The argument that openness inherently risks safety is a form of semantic gymnastics used to justify building opaque, black-box models that monopolise control and knowledge,” he added.

Instead, transparency is woven directly into ROMA’s architecture. Because tasks propagate through an explicit recursive tree, every stage is traceable, auditable, and open to inspection. Builders can review how tasks were broken down, how context flowed, how tools were invoked, and where failures happened, something impossible in monolithic models.

ROMA is fully open-source and available for inspection and modification, reinforcing Sentient’s belief that the future of AI must be democratic, not centralised.

The Global, Not Local, Roots of ROMA

Although India features prominently in Sentient’s leadership and mission, ROMA’s design philosophy is not geographically bound. “Our team, researchers and vision are fundamentally global… AI should be borderless and not controlled by any single entity or sovereign state,” Tyagi clarified.

This borderless ethos directly shaped ROMA’s modular architecture. Contributors can plug in models, tools, datasets, or agents without relying on proprietary backends or closed APIs. More importantly, Sentient’s ecosystem is designed to reward contributors, not just extract value from them.

As Tyagi put it, ROMA’s architecture is meant to “provide the necessary architectural openness while also offering a transparent, traceable meta-agent framework that allows this global talent pool to build and control their own sovereign AI models.”

In a world where AI infrastructure threatens to be monopolised by a handful of technology giants, the open-source model offers not just an alternative, but a counterforce.

Why Developers Are Paying Attention

ROMA represents the next stage in a broader mission for Sentient co-founder Sandeep Nailwal, who previously helped with Polygon’s decentralisation of finance.

“With Sentient, we’re doing the same for intelligence. We want to stop AI from becoming a gated, private resource controlled by a handful of companies. Intelligence should be a public good,” he told AIM.

Just as Web3 democratised access to financial infrastructure, ROMA aims to democratise access to cognitive infrastructure. If Polygon built the rails for decentralised value, Sentient seeks to make the rails for decentralised intelligence.

Nailwal views this as critical for global equity. “If it’s locked away, it kills participation… From startups to sovereigns, everyone needs access to intelligence as a public utility,” he said.

The company’s $1.2 billion valuation, as Nailwal explained, is anchored in investor confidence that open-source innovation scales faster and more efficiently than closed systems.

“No single company, no matter how large, can out-innovate a global network of open builders moving in parallel,” he clarified.

A key component of this ecosystem is GRID, Sentient’s open intelligence marketplace, where developers can contribute tools, models, and agents and earn based on their usage. This solves one of the fundamental challenges of open source: sustainability.

More importantly, the open source framework offers a philosophical and structural alternative to the consolidation of AI power. “Sentient represents Web3 for intelligence… We’re decentralising AI itself,” Nailwal summarised.

The startup said in its blog that ROMA’s prototype search agent, dubbed “ROMA Search”, achieved 45.6% accuracy on the SEALQA subset known as “Seal-0,” which tests complex multi-source reasoning, outperforming the previous best system, Kimi Researcher, at 36% and the proprietary Gemini 2.5 Pro at 19.8%.

The post Sentient Pushes Open AI Infrastructure into the Global Spotlight appeared first on Analytics India Magazine.

Google has been relatively quiet when it comes to making agentic coding or vibe coding announcements. The furthest it went was with Jules, Vertex AI, or Gemini CLI, obviously apart from acquiring Windsurf. But now, the company has decided to enter the field formally with Antigrativy.

The company is trying to land its agentic IDE into a space packed with hype, skepticism, forking drama and an audience tired of experimenting with half-finished tools. Antigravity allows agents to “autonomously plan and execute complex, end-to-end software tasks” with direct access to an editor, terminal and browser.

It comes with Gemini 3, which is already ruling a ton of hearts amongst creators and developers in Cursor, GitHub and Replit.

Out in the open, opinion on Antigravity remained split. Some are cancelling their $6o Cursor subscriptions, some claim “Google is changing the VIBE CODING game.” Others say that it seems like it’s still in Beta and is worse than Cursor and Copilot.

But What’s Really Happening?

Coming to the positives first, Antigravity does things that others cannot. It can do full screen recordings to verify your app actually works in real time. “Screen recording + live debugging gives AI the kind of context developers used to dream about,” a software engineer demoed on X.

I’m bullish on Google Antigravity for 1 simple reason.
It can now do full screen recordings to verify your app actually works in real time.
Screen recording + live debugging gives AI the kind of context developers used to dream about. pic.twitter.com/q64kVVyNjO

— corbin (@corbin_braun) November 20, 2025

Some are obsessed with the UI, while others with the agentic capabilities, where developers were able to fix errors that other AI editors struggle with. Some are in love with how it integrated Nano Banana Pro for multimodal reasoning.

But for many long time AI coding tool users, Antigravity arrived with a shape most developers recognised in the first second. A Hacker News thread on launch day saw more than a thousand comments, with users repeating a familiar line.

While one person wrote: “Oh no. Not another VSCode fork…,” another said, “Oh cool, another IDE for programming… aaaand it’s a vscode fork. I don’t know what I expected tbh.” This captured the fatigue that has grown around the rise of agentic editors that look the same, act the same, and often stumble the same.

Yet, this launch was not only about VS Code. In an attempt to push out Cursor, Antigravity pushed an older story back into view: Google’s decision in July 2025 to hire Varun Mohan from Windsurf and license its technology for roughly $2.4 billion. The Windsurf acquisition by Cognition Labs came later, though many developers now believe Antigravity carries Windsurf’s fingerprints, with Mohan in the lead.

This is reflected in its references to Cascade, the proprietary agent system inside Windsurf, which were spotted inside Antigravity’s code. Visual Studio Magazine had already hinted at the same direction in its piece titled “Google Joins AI IDE Race to Compete with VS Code, Apparently Forking VS Code.”

And yes, the discussion continues.

Rate Limits the Real Issue, At Least for Now

Adarsh Shirawalmath, founder of Tensoic, a company that provides custom LLMs, fine tuning, and inference, told AIM Antigravity provides good quality edits and debugging/testing tools in the browser, and brought new ideas that stood out. “It’s the first good agentic IDE I’ve used that can actually connect to my remote SSH machine through the CLI, run commands there, manage dependencies, and navigate the entire remote filesystem properly.”

He said it “may follow some design aspects of Windsurf but it is a form of VS code.” Shirawalmath was clear about the flaws too. “There are quite a lot of bugs such as agents being stuck in a loop, other models not working, quicker rate limits consumption.”

This is a recurring theme on discussion forums as well. “There were some UI glitches,” the top commenter on Hacker News said. Cursor, he argued, had “real annoying usability issues” and he found Antigravity “more polished.” He imported his old settings, got working on a project, enjoyed the Gemini 3 model for a while and then hit a wall.

But then, the dreaded rate limits kicked in. “After about 20 mins – oh, no. Out of credits,” he said, while adding that he looked for a purchase button, but found none. “If you release a product, let those who actually want to use it have a path to do so,” he wrote.

He switched back to Cursor soon after and discovered that Cursor itself already had Gemini 3 Pro. His conclusion was brutal: “Real developers want to pay real money for real useful things.” This reminded of a similar scenario when developers were cancelling Cursor subscriptions.

Adithya S Kolavi, founder of CognitiveLabs, told AIM that he has been trying to use Antigravity, but the rate limits don’t allow him to. “Not an overall good experience so far,” he said, while adding that integrating into the browser is unique. “Like most IDEs just focus on text basically code, they [Google] have taken a more multimodal approach natively, which is nice,” he said.

On the company side, Google tried to calm the situation. Varun Mohan said, “We’re also aware of the capacity constraints everyone is facing given our growth and working to address them as quickly as we can,” he said, which is fair since it is still in beta.

But, the forking allegations continue as some developers call it Windsurf 2.0 with a new UI.

wow google antigravity is a vscode fork with a seperate agent view ui
this is hilarious

— dax (@thdxr) November 18, 2025

This is where the story rests today. Antigravity has people excited, irritated, curious, hopeful, and tired — all at once. Google has built something that feels half Windsurf, half VS Code and fully trapped inside the expectations of a community that has seen too many promises from too many AI IDEs.

The post Why Developers are Fighting Over Google’s Cursor Killer Antigravity appeared first on Analytics India Magazine.

As enterprise IT evolves, the traditional model of managing IT services is being rewritten. What was once a service-heavy, project-led function is now morphing into a platform-driven, software-led ecosystem. This shift towards service-as-a-software, is redefining how Indian IT companies deliver, consume, and also compete in IT Service Management (ITSM).

The global IT Service Management (ITSM) market is consolidating into a platform-led ecosystem powered by AI, automation, and cloud-native operations. Valued at nearly US $12 billion in 2024, it is projected to exceed US $36.78 billion by 2032, growing at over 15% annually.

ServiceNow dominated with about 51.1% share in 2023, setting the standard for workflow automation, while Salesforce, BMC, and Atlassian are expanding their footprints through AI-integrated service layers.

The ITSM space is witnessing strategic partnerships that blur the line between software vendors and IT service firms. Salesforce is emerging as a challenger with its Agentforce IT Service, backed by an HCLTech alliance for AI-led enterprise solutions. BMC has teamed up with Infosys and Tech Mahindra on hybrid-cloud and automation, while ServiceNow has collaborated with Wipro, Tech Mahindra, and NTT DATA.

Service-as-a-software

Achyuta Ghosh, executive research leader HFS Research, says this is the Services-as-Software shift in motion. “As ITSM becomes productised, IT vendors are reinventing themselves from service integrators to platform providers.” They’re moving from time-and-material models to subscription-led outcomes, monetising uptime, resolution speed, and productivity rather than manpower,” he adds.

This evolution turns IT services into software-powered assets — repeatable, scalable, and designed for measurable business value.

The future of ITSM lies in consuming intelligent service platforms as a product, not as a project, according to Shelton Rego, VP of India business at Freshworks, a company that provides enterprise-grade service software solutions.

He says that modern businesses are looking for faster results, instead of slow rollouts, and uncomplicated AI-powered solutions, instead of consulting contracts. “Software-led servitisation delivers that promise, transforming ITSM into an engine of growth and efficiency. Vendors who build for speed, simplicity, and real-time adaptability will define the next era of service management,” Rego adds.

Therefore, companies providing ITSM remove complexity, so that organisations can focus their energy on customers and business outcomes rather than wrestling with IT processes.

How are IT companies adopting?

Across IT functions, adoption is accelerating as AI and automation embed themselves into daily workflows. Platforms like ServiceNow have redefined how IT teams handle incidents, changes, and problems.

Vineet, an IT service delivery professional at a large IT company, who wishes to go by his first name, explains: “Earlier, teams used multiple tools like Outlook for notifications, separate dashboards, and manual reporting. With ServiceNow, all of that happens in one place.” Built-in dashboards now visualise real-time ticket data, track change progress, and automate stakeholder alerts.

ServiceNow’s GenAI capabilities have further simplified operations. Based on historical incidents, the platform can generate proactive recommendations and summaries reducing reliance on separate knowledge databases. “Earlier, we had to use different tools for root cause analysis, but now GenAI does it automatically inside ServiceNow,” Vineet adds.

However, not all functions are fully automated. Post Implementation Reviews (PIRs) and final change reviews still require manual effort. “ServiceNow hasn’t yet matured in that area,” he says.

This mix of automation and manual intervention defines the current state of IT adoption rapidly evolving, but still balancing between machine intelligence and human oversight.

Business model

The platformisation of ITSM is not only a technology shift but also a business transformation. As automation and AI integrate deeper into delivery, IT firms are moving from manpower-based contracts to software and subscription-led engagement.

Ghosh explains, “Vendors are developing proprietary IP, automation frameworks, and cloud-native platforms that can be licensed or co-delivered with hyperscalers. Instead of billing for effort, they’re monetising outcomes—like uptime and productivity gains.”

This transformation is also changing workforce composition. Rego emphasises a “people-first AI strategy” where repetitive tasks are automated, allowing talent to focus on creative and strategic roles. “Traditional headcount-driven models are being replaced by AI strategies which focus on automating the mundane tasks and upskilling talent to do the things that only humans can do,” he says.

For IT service majors, this means balancing two operating models: one focused on scalable software products and another on managed services that integrate those products into enterprise ecosystems.

Integration Challenges

One of the primary challenges in transforming ITSM is integrating legacy systems that were not built for agility.

Regosays that these legacy systems often lack the flexibility to support modern day DevOps practices, leading to compatibility issues and slower deployment cycles. “Cloud and AI-native solutions are built with platform capabilities in mind. Unified platforms like Freshworks which have ITSM, ITAM and ITOM capabilities make it easier for companies to not just onboard quickly but also to start realising return on investment much faster,” he says.

Ghosh adds, “Transforming ITSM from process to product isn’t easy. Vendors face challenges such as standardising highly customised processes, integrating with legacy backends, and building unified data models for automation.”

Codifying ITSM requires profound architectural change—APIs, low-code frameworks, and AI-driven decisioning, not just process digitisation, he says, adding that IT service teams must re-skill from process managers to software engineers, adopting agile release cycles and product thinking.

Convergence creates competition

The ITSM market’s evolution is creating new competitive pressures. As SaaS companies encroach on areas once dominated by IT service firms, both are converging around the same goal—automation-led, software-powered service delivery.

Ghosh observes that the rise of ITSM-as-software blurs traditional boundaries between IT service providers and SaaS players. “While today they collaborate with services firms acting as implementation partners, over time, both will compete for the same automation and operations budgets.”

In India, this is reshaping customer behaviour, as they seek cost-effective digital solutions with speed and simplicity. “Our clients typically go live in under 90 days, and AI solutions in under 30 minutes. This rapid time-to-value is putting pressure on traditional IT service providers to adapt or risk obsolescence,” says Rego.

Ghosh says that the decision to go for an IT company or SaaS vendor depends on enterprise maturity. Digital-native organisations increasingly adopt ITSM platforms directly, valuing agility, control, and transparency. In contrast, large traditional enterprises still rely on IT service providers to manage ITSM due to integration complexity, legacy systems, and hybrid operations, he says.

However, competition is not only between vendors—it’s also driven by enterprise preferences. As Vineet explains, clients dictate tool choices. “Even if we have our own AI-enabled ITSM tools, customers often insist on ServiceNow or Salesforce. They know these are market leaders and prefer not to take risks with alternatives,” he says.

This customer-driven consolidation means newer or niche vendors must differentiate through speed, AI integration, and ROI visibility rather than tool variety.

ITSM as a platform marks a shift from manpower-driven services to software-led, AI-powered platforms. As IT firms and SaaS players converge, success will depend on delivering speed, simplicity, and measurable outcomes. The future of ITSM lies in turning service delivery into a strategic, software-defined advantage for enterprises.

The post Indian IT Firms Race Ahead with Software-Led, AI-Powered ITSM appeared first on Analytics India Magazine.

India’s healthcare market, estimated anywhere between $180 and $400 billion, is undergoing a significant shift, and digital health is now one of its fastest-growing engines. Startups are playing a pivotal role in this transformation.

Over the past few years, India’s health-tech segment has expanded from around $3 billion in 2020 to $7 billion in FY 2023, with projections placing it at a massive $60 billion by FY 2028. The ecosystem itself is booming: India now hosts over 10,000 health and life sciences startups recognised by promotion of industry and internal trade department (DPIIT), recording a remarkable 127% CAGR since 2016.

Against this backdrop of scale and innovation, here are the top 10 Indian healthtech startups leading the race.

Niramai
Niramai offers Thermalytix, a non-invasive, radiation-free breast cancer screening system using thermal sensors. Its AI analyses ~4,00,000 temperature points per scan to detect subtle heat signatures associated with tumours, producing a quantitative ‘breast health score’. Their models are trained on large datasets linked to mammograms, ultrasound and histopathology, increasing clinical reliability. Thermalytix is now commercially available in over 22 countries and holds a CE mark.

Qure.ai
Qure.ai builds AI-powered diagnostic tools for medical imaging, including chest X-rays (qXR) and head CT scans (qER), to detect conditions like tuberculosis, lung nodules and haemorrhages. Their deep learning models speed up diagnosis by triaging critical cases in real time, reducing radiologists’ workload. Recently, Qure.ai announced plans to expand to 10,000 hospitals in the coming years.

SigTuple
SigTuple combines robotics and AI via its AI100 platform to digitise microscope slides, including blood smears and urine, and analyse them. Their AI models classify different cell types, detect morphological anomalies and flag abnormalities. The AI100 product is FDA 510(k) approved, and SigTuple holds multiple patents on its AI-based screening tools.

Tricog Health
Tricog Health offers remote cardiac diagnostics. ECGs captured in clinics are uploaded to the cloud, where Tricog’s AI analyses them for over 140 cardiac conditions, like arrhythmias and ST elevation. Its AI-generated report is reviewed by in-house cardiologists to ensure accuracy and speed. As of late 2024, Tricog has enabled diagnostic care for 20 million lives, partnering with organisations like AstraZeneca to scale in underserved regions.

HealthifyMe
HealthifyMe is a digital wellness platform that combines tracked nutrition, fitness and lifestyle with human coaches. Its AI coach ‘Ria’ provides personalised diet and advice on habits, and the company is now embedding generative AI to personalise nutrition plans further. In its latest funding round, it raised $30 million to expand AI capabilities and push into global markets.

Practo
Practo provides a comprehensive healthcare platform for patients and clinics: booking appointments, managing medical records and running clinic software. Their AI supports smart triage, structured patient summaries, and optimised scheduling. Practo is now aiming to double its international revenue (currently ~20%) by expanding into markets like Canada and Australia.

PharmEasy
PharmEasy is a leading Indian e-pharmacy that also offers diagnostic test bookings. While specific details on AI deployment are not widely publicised, it leverages machine learning for demand forecasting, inventory optimisation and personalised test-and-medicine bundling. The operational efficiency gains from AI help reduce costs, improve delivery times and ensure better availability.

Innovaccer
Innovaccer provides a data platform that aggregates fragmented health records into a unified, AI-enabled system. Their AI predicts care gaps, stratifies patients by risk, automates documentation via NLP, and supports decision-making for care teams. In January, they raised $ 275 million (Series F) to build new AI co-pilots and scale cloud capabilities.

MedGenome
MedGenome is a genomics company offering genetic testing, sequencing, and bioinformatics. Their AI-driven pipelines interpret genetic variants predict disease risk, and support personalised medicine. Recently, they expanded in India by acquiring a majority stake in Gujarat-based diagnostic chain Green Cross Genetics to strengthen their integrated genomics and diagnostics footprint.

Tata 1 mg
Tata 1mg is a full-stack digital healthcare platform offering e-pharmacy, diagnostics, teleconsultations, and wellness content. Since being acquired by Tata Digital, it has built a tighter supply chain stocking medicines in its own warehouses, which has boosted margins and improved quality control. On the tech side, it uses data and AI in its ODIN (Order Delivery Intelligent Network) to manage orders, inventory, quality checks and logistics. ata 1mg is now aggressively expanding offline, aiming for omnichannel reach.

Disclaimer: This list is not a ranking and is not based on any specific metric. The startups featured here are highlighted for their contributions to India’s healthtech ecosystem and their use of AI, not for comparative evaluation.

The post Top 10 Indian Startups Powering Healthcare with AI appeared first on Analytics India Magazine.

The United States Court of Appeals for the Fifth Circuit has upheld the damages imposed on Tata Consultancy Services (TCS) in the trade secrets lawsuit filed by DXC Technology Company, earlier known as Computer Sciences Corporation (CSC).

In a disclosure, TCS said it is evaluating various options, including further review and appeal before the appropriate courts, and intends to “vigorously defend its position.”

The Appeals Court has also set aside an earlier injunction issued by the United States District Court for the Northern District of Texas.

That injunction had restricted TCS from using certain CSC software and confidential materials. With the order now vacated, these restrictions are no longer in force while the District Court reassesses the matter.

A major part of the dispute concerns TCS’s work for Transamerica.

Under Transamerica’s agreements with CSC, TCS was given limited permission to access CSC’s Vantage and CyberLife systems to support the insurer’s technology transformation.

CSC later argued in court that TCS went beyond what was allowed and used CSC’s software and confidential information in ways not permitted under those agreements.

The District Court said CSC had raised genuine issues of fact on this point, allowing the claim to proceed.

The original judgement had found TCS liable under the Defend Trade Secrets Act of 2016 and ordered the company to pay CSC about $56.15 million in compensatory damages, $112.30 million in exemplary damages, and $25.77 million in prejudgment interest through June 13, 2024.

TCS disclosed this ruling last year, saying it believed it had “strong arguments against the Judgement” and intended to pursue review or appeal.

The company also said the ruling would not have a major adverse impact on its financials or operations.

The District Court’s detailed memorandum opinion from 2023, had denied TCS’s motion for summary judgment and its attempt to exclude CSC’s damages expert.

It also struck several of TCS’s defenses, including failure to state a claim, lack of trade secrets and consent, and ruled in CSC’s favour on defenses such as laches, failure to mitigate and unclean hands.

Other defenses, such as equitable estoppel, waiver and acquiescence, were left to be decided at trial.

TCS had also argued that the exemplary damages were “legally excessive” and should be reduced or set aside.

The Appeals Court rejected this argument and upheld the entire damages awarded by the District Court.

The post US Court Upholds $194 Million Penalty Against TCS appeared first on Analytics India Magazine.

Salesforce CEO Marc Benioff has sparked a fresh wave of debate in the AI world after declaring that Google’s newly launched Gemini 3 has decisively overtaken ChatGPT.

In a post on X, Benioff wrote that he has used ChatGPT “every day for 3 years,” but after spending two hours with Gemini 3, he’s “not going back,” calling the leap in reasoning, speed, images and video “insane.”

Benioff’s emphatic endorsement comes just as Google unveiled Gemini 3, its most powerful multimodal AI model to date and Nano Banana Pro, an advanced image-generation system built on top of it.

His comments immediately intensified comparisons between Google and OpenAI, the two companies locked in the most closely watched AI rivalry.

According to Google, Gemini 3 brings major improvements in complex reasoning, multimodal understanding, and tool-use capabilities. It integrates text, images, video, and code processing, positioning it as Google’s first truly general-purpose, agentic AI system across consumer and enterprise products.

Alongside it, Google introduced Nano Banana Pro, a new image-generation and editing model that promises studio-grade visuals, reliable text rendering, multilingual support, consistent brand styling, and high-resolution (including 4K) output.

The model is already rolling out across Google Workspace and the Gemini app, signalling Google’s push to tie creative workflows directly into its AI ecosystem.

The launches represent one of Google’s most aggressive bids yet to reclaim AI leadership, especially as OpenAI continues rapid advancements with its GPT-5 series. Benioff’s praise adds fuel to that momentum, offering rare public validation from a long-time power user of competing AI systems.

The post ‘Holy S***… I’m Not Going Back to ChatGPT,’ Says Marc Benioff After Using Gemini 3 appeared first on Analytics India Magazine.

Andrej Karpathy, the AI researcher and founder of Eureka Labs, recently shared an experiment called “LLM-Council”, which sends a user query to multiple language models, lets them anonymously judge each other’s answers, and then produces a final response based on their rankings.

The results of this experiment revealed that the AI model that consistently ranked highest was OpenAI’s GPT-5.1. This is significant given how recent benchmarks suggested Google’s Gemini 3.0 had overtaken OpenAI in overall capability and reasoning tests.

“Quite often, the models are surprisingly willing to select another LLM’s response as superior to their own, making this an interesting model evaluation strategy more generally,” said Karpathy.

“For example, reading book chapters together with my LLM Council today, the models consistently praise GPT 5.1 as the best and most insightful model, and consistently select Claude as the worst model, with the other models floating in between.”

Karpathy’s experiment setup is a three-step loop.

First, the user’s query is sent to all models separately, and their answers are shown side-by-side without revealing who wrote what.

Next, each model sees the others’ responses, still anonymised, and ranks them based on accuracy and insight. Finally, a “chairman model” produces the answer by combining the councils’ outputs and their critiques, turning the response into a consensus built through competition.

However, Karpathy also noted that these rankings are subjective and don’t necessarily match his own judgment.

As he put it, “I’m not 100% convinced this aligns with my own qualitative assessment. For example, qualitatively, I find GPT 5.1 a little too wordy and sprawled and Gemini 3 a bit more condensed and processed. Claude is too terse in this domain.”

He revealed that he built this project over the weekend using a ‘vibe coding’ tool and shared the repository on GitHub.

Reacting to Karpathy’s post on X, Vasuman M, founder and CEO at Varick AI Agents, claimed on the social media platform that he built something similar months ago, and observed similar performance from OpenAI’s models.

“Even after plugging in Gemini 3.0, the winner was GPT 5.1, every single time,” he said. “Even funnier, if you tell other models (Claude, Gemini, Grok) that the answer they are reading came from GPT (un-anonymise), they fold immediately and start correcting themselves based on GPT’s output.”

The post GPT 5.1 Is The Best, As Declared By Gemini 3.0, Claude & Grok On Andrej Karpathy’s ‘LLM Council’ appeared first on Analytics India Magazine.