DEV Community: Thirumalaivasan GJ

Laptop Ultra for developers: what the RTX Spark chip actually gives you

Thirumalaivasan GJ — Sun, 15 Nov 2026 05:12:35 +0000

Late on this one. Microsoft showed the Surface Laptop Ultra at Computex on June 1 and opened pre-orders on Oct 7.

The part that matters for devs is the chip. It's Nvidia's RTX Spark platform, an Arm CPU plus a Blackwell RTX GPU sharing unified memory.

Specs:

Top config: 20-core CPU, 6,144 GPU cores, up to 128GB unified memory
Second config: 18-core CPU, 5,120 GPU cores, up to 64GB
Up to 1 petaflop of FP4 AI performance (Microsoft's claim)
CUDA support, and Microsoft says it can run local agent workflows with models up to 120B parameters
15-inch mini-LED touchscreen, under 4.5 lb, USB-C, USB-A, HDMI, SD card reader

Price starts at $2,599 (18-core, 24GB RAM, 512GB) and goes to $5,899.99 for the 128GB model, per Notebookcheck. US pricing only. No India launch announced.

Caveats before you get excited:

It's Windows on Arm, so x86 apps and games depend on emulation
The MacBook Pro speed comparisons are Microsoft's own numbers
"All-day battery" is a claim, and Microsoft says it varies a lot
How the RTX hardware fits into Copilot+ PC features is still unclear

If you run big local models, it's interesting. If you build web apps and use Docker, it's probably overkill.

Sources:

Notebookcheck: https://www.notebookcheck.net/Microsoft-Surface-Laptop-Ultra-debuts-with-RTX-Spark-128-GB-unified-memory-and-2-599-starting-price.1418420.0.html
Windows Central: https://www.windowscentral.com/hardware/surface/microsoft-surface-laptop-ultra-announced-computex-2026
Tom's Hardware: https://www.tomshardware.com/laptops/microsoft-and-nvidia-launch-surface-laptop-ultra-with-rtx-spark-rtx-spark-preorders-live-now-coinciding-with-major-windows-11-changes-for-agentic-ai
Microsoft: https://microsoft.com/ultra

Self-hosted Claude Managed Agents: what you run, what still leaves your network

Thirumalaivasan GJ — Sat, 14 Nov 2026 12:31:31 +0000

Anthropic opened a private beta of self-hosted Claude Managed Agents on Nov 12. Access is by request through your Anthropic account team.

Before this, you could self-host the sandbox where an agent's code runs, but Anthropic ran the rest, including credentials and the record of what each agent did. Now the whole service can sit in your cloud or on-prem.

What you operate:

the Managed Agents server
its PostgreSQL database (session history lives here)
the workers where agents run shell and file tools

You build agents with the same core APIs as the hosted service.

What still leaves your network is Claude itself. Model requests go where you point them: Anthropic, Amazon Bedrock, Google's Agent Platform, Microsoft Foundry, or your own gateway. Prompts out, completions back. Zero data retention agreements cover those requests like any other Messages API call.

Security bits worth knowing:

Agent loop, session history and vault secrets sit outside the sandbox, and credentials stay out of it too
A proxy outside the sandbox attaches credentials for MCP tools and supported API calls, and you can keep them in your own secrets manager
That covers MCP tools and supported API calls only, so check any other service your agent reaches
Per-environment host allowlists block everything else, and when you host the sandbox the limits live in your firewall
Every tool call and result is logged and exports over OpenTelemetry

The cost is ops work. Patching, securing and configuring the server, database and firewall is on you. The Next Web also reports nothing migrates from the hosted service and webhooks and scheduled runs are missing for now (single outlet, check the docs).

Sources:

Anthropic: https://claude.com/resources/articles/self-hosted-claude-managed-agents
Help Net Security: https://www.helpnetsecurity.com/2026/11/12/anthropic-self-hosted-claude-managed-agents/
The Next Web: https://thenextweb.com/news/anthropic-self-hosted-claude-managed-agents

Cognizant is putting 25,000+ people on Claude. Here's what's confirmed

Thirumalaivasan GJ — Sat, 14 Nov 2026 05:31:40 +0000

Cognizant announced on Nov 10 that it's committing more than 25,000 professionals to a new unit, the Claude Frontier Capability Group. The job is helping enterprise clients put Anthropic's Claude into production.

CEO Ravi Kumar S said "we are putting more than 25,000 of our people on Claude." Anthropic's chief commercial officer Paul Smith described the same thing as Cognizant "committing more than 25,000 people to help their customers put Claude to work."

Alongside the unit, Cognizant introduced a managed services offering. Cognizant runs and maintains Claude-powered systems for clients. The package combines dedicated staff, licensed Claude access and Cognizant's own software, including something it calls its AI Delivery Operating System. The pitch is more predictable AI spend plus security and governance controls. No commercial terms were disclosed.

Other numbers from the announcement:

44,000+ Cognizant employees trained or certified on Claude
26,000+ Claude certifications across the company
Claude-powered agents will be used to modernize legacy and mainframe systems

Cognizant also shared client results, like a pharma manufacturer cutting troubleshooting from 90-120 minutes to 10-15, and an insurer speeding up some operational processes 40-60%. Those figures are Cognizant's own and haven't been independently verified.

One thing worth saying plainly. Some coverage frames this as Cognizant moving away from traditional engineer staffing. I couldn't find that in the announcement or the reporting. What's confirmed is a large dedicated Claude team and a managed services model. Whether it changes how Cognizant staffs everything else is an open question.

For devs, the practical read is that Claude, MCP and agent tooling are becoming a standard skill in services work, and Cognizant isn't the only one. HCLTech signed its own Anthropic partnership in October, per Analytics India Magazine.

Sources:

Analytics India Magazine: https://analyticsindiamag.com/ai-news/cognizant-commits-25000-professionals-to-deploy-claude-in-enterprises
Yahoo Finance (press release): https://finance.yahoo.com/technology/ai/articles/cognizant-establishes-25-000-person-151221219.html
BW People: https://www.bwpeople.in/article/cognizant-deploys-25-000-professionals-for-anthropic-s-claude-633398

Microsoft 365's Nov 12 outage: what's known and what developers can take from it

Thirumalaivasan GJ — Fri, 13 Nov 2026 13:53:50 +0000

Microsoft 365 had a major outage on November 12, 2026. This post covers what's confirmed from IT-Connect, Astra Security and Euronews, and ends with some general takeaways that are my own and not from Microsoft.

What happened

Started around 17:54 UTC on November 12 and was largely resolved around 05:33 UTC on November 13, per Astra. Reports of the duration vary between about 10 and 12 hours.
Affected Exchange Online (Outlook on all platforms, including REST API access), Teams, Copilot, SharePoint Online, OneDrive, Intune, Purview and Defender. On-premises Exchange and SharePoint were not affected (Astra).
Downdetector showed nearly 130,000 Outlook reports at the peak and users in 74 countries (IT-Connect).

Microsoft's explanation

First: "higher than normal load" on part of its infrastructure.
Later: not a traffic spike. Servers became unavailable and failed health checks, and a process that establishes secure connections became overwhelmed.
Astra describes the remaining servers taking the extra load and failing too. That reading is Astra's interpretation of Microsoft's updates.
Microsoft began rolling back a recent network configuration change that matched when impact started, and said it knew of no malicious activity (IT-Connect).

Restoration

Microsoft validated an alternative health check method and gradually reintroduced servers. It warned servers could become unresponsive again under traffic (Astra). A root cause report hadn't been published when I checked.

Downstream impact

Astra says OpenAI and Anthropic connectors that depend on Microsoft APIs reported related errors. I only found that in one source.

General takeaways (mine, not Microsoft's)

Health checks that remove servers can raise load on the rest, so design them with care.
If your app depends on Exchange or Graph APIs, plan for degraded behavior and not only for failure.
Subscribe to your providers' status feeds and have a fallback message for users.

Sources: IT-Connect, Astra Security, Euronews

Gemini 4 Hydrogen for developers: pricing, thinking levels and what's verified

Thirumalaivasan GJ — Fri, 13 Nov 2026 05:33:36 +0000

Google released Gemini 4 Hydrogen on November 12, 2026. Here's what matters if you build with the API, based on Google's blog post and PPC Land's coverage. Every benchmark below is Google's own and, per PPC Land, hasn't been independently verified.

What it is

Google calls it "our daily driver for agentic workflows." It sits below Gemini 4 Carbon and Argon, which handle harder reasoning.

Thinking levels

There are four: low, medium, high (the default) and a new xhigh. You pick the reasoning depth per task, so you wait and pay only for what you need.

Speed

At the default high setting Google says Hydrogen finishes tasks about 2.5x faster than Gemini 3.8 Flash. Some enterprise partners saw 2x to 2.8x speedups with 12% to 23% lower token use.

Pricing

Input: $1.50 per 1M tokens
Output: $7.50 per 1M tokens

It's flat across the 1M token context window. PPC Land says that's 25% below Argon's introductory rates.

Benchmarks (Google's numbers)

Artificial Analysis Intelligence Index: 51 at high (Argon 50, Carbon 59), 54 at xhigh
Terminal-Bench 4.1: 57% vs 18.2% for 3.8 Flash
Vals Tax Agent Bench: 52.3%, ranked first
Vals Legal Research Bench: 55.8%, tied for first

Google uses Terminal-Bench 4.0 in some comparisons and 4.1 in others, so check which one you're reading.

Where to use it

Gemini API, Google AI Studio (all tiers) and Android Studio
Gemini for business, Antigravity and Gemini in Google Sheets for eligible business plans
Antigravity access is limited to paid Google AI Pro and Ultra subscribers
Gemini app (free users get it through automatic model selection) and Google Search for all users

What I couldn't confirm

Some outlets say Hydrogen retires the Flash brand. Google's blog doesn't say that. I'd also run your own evals before switching a production workload, since the numbers above are vendor-reported.

Sources: Google blog, PPC Land

Building a Chennai Metro Rail Travel Planner During My CMRL Internship 🚇

Thirumalaivasan GJ — Thu, 12 Nov 2026 14:03:18 +0000

Building a Chennai Metro Rail Travel Planner During My CMRL Internship 🚇

Public transportation becomes much easier when people can find the information they need without jumping between multiple sources.

During my internship at Chennai Metro Rail Limited (CMRL), I got the opportunity to work with a team on a project that addressed this problem: the Chennai Metro Rail Travel Planner.

In this post, I want to share a little about the project, the work I contributed, and what I learned from the experience.

What is the Chennai Metro Rail Travel Planner?

The Chennai Metro Rail Travel Planner is a web application designed to help commuters plan their metro journeys.

Instead of manually checking different sources, users can select their starting station and destination to get useful travel information in one place.

Some of its key features include:

Selecting origin and destination stations
Viewing available metro routes
Checking estimated travel fares
Viewing journey duration and distance
Exploring intermediate stations
Checking train schedules and upcoming trains
Exploring routes through an interactive metro map

The goal was to make journey planning simpler and more convenient for commuters.

What did I work on?

During my internship at CMRL in May–June 2024, I worked with the team on the Travel Planner, primarily focusing on the frontend and user experience.

My contributions included:

Working on the frontend interface and layout
Improving the user experience and responsive design
Integrating interactive map features using Leaflet
Working with JavaScript to handle user interactions
Connecting the frontend with the available Chennai Metro APIs

One part I particularly enjoyed was working with the interactive map. It made the application more useful by allowing commuters to explore stations and understand their routes visually.

The backend and internal services were handled separately, so the project was a collaborative effort.

Technologies I worked with

The project gave me practical experience with:

HTML & CSS – Building and styling the interface
JavaScript – Handling interactions and application logic
Leaflet.js – Creating interactive maps
Chennai Metro APIs – Working with metro-related information

It was a good opportunity to apply what I had learned outside the classroom and understand how different parts of a real-world application come together.

What I learned

This experience taught me that building an application is not just about writing code.

A useful application also needs a clear interface, reliable information, and a smooth experience for the people using it.

Working on a project connected to public transportation also gave me a different perspective. Even small improvements in how information is presented can make an application easier to use.

Most importantly, I got to experience working on a real-world project as part of a team.

Final thoughts

My internship at Chennai Metro Rail Limited was a valuable learning experience, and the Travel Planner remains one of the projects I'm happy to have contributed to.

It's always satisfying to see something you've worked on being used for a practical purpose.

Project: https://travelplanner.chennaimetrorail.org/

I'll be sharing more about the technical side of this project and other things I've worked on in future posts.

Thanks for reading!

Hey DEV Community! I'm Thirumalaivasan GJ👋

Thirumalaivasan GJ — Thu, 12 Nov 2026 13:52:20 +0000

Hey DEV Community! I'm Thirumalaivasan GJ

Hey everyone!

I'm Thirumalaivasan, a Computer Science graduate from SRM Institute of Science and Technology, India.

I enjoy building things, exploring new technologies, and occasionally going down rabbit holes trying to understand how something actually works.

Over the past few years, I've explored different areas of software development, especially Full-Stack Development, Generative AI, Cloud, and DevOps.

I've had the opportunity to intern at Chennai Metro Rail Limited and ITC Cloud Tech. One experience I particularly enjoyed was contributing to the Chennai Metro Rail Travel Planner, a live application built to help commuters plan their journeys.

I've also worked on projects like CivicGPT, an AI-powered tax guidance system using RAG, and experimented with Kubernetes, automation, and cloud technologies.

When it comes to technology, I spend most of my time exploring:

Python and JavaScript
React and FastAPI
Generative AI and RAG
Docker and Kubernetes
AWS and CI/CD

What brings me to DEV?

I've been exploring different developer communities, reading technical discussions, and sharing things I've learned.

I already use platforms like Reddit and Medium, and now I wanted to spend some time here on DEV as well.

I'd like to share more about the things I build, technologies I experiment with, problems I run into, and lessons I learn along the way.

I'm not here pretending to know everything. There's always something new to learn, and that's probably what I enjoy most about this field.

Looking forward to connecting with you all!

If you're working on something interesting, feel free to say hello.

Find me here:

GitHub: https://github.com/Thirumalaivasangj3
LinkedIn: https://www.linkedin.com/in/thirumalaivasangj3
Portfolio: https://thirumalaivasangj.vercel.app/