Mailr — Monday, 31 August 2026
 The AI Brief for India's Tech Builders
Monday, 31 August 2026
|
|
|
A rakhi is a promise of protection with no terms attached. Fitting, then, that 130 companies just signed an open letter urging a surge in cyber defence. OpenAI, Anthropic, Google, Visa, Citi, General Motors — all on the same page. Nobody committed to anything. Everyone tied the thread.
Here's your Mailr:
|
|
|
This could be you |
|
Your product, in front of 200+ tech builders across India. Tap the logo to contact us
|
|
|
|
|
🌍 Global AI |
OpenAI, Anthropic and Google jointly call for a cyber defence surge |
 Image source: Generated using Google Gemini
|
| Over 130 organisations, including direct rivals OpenAI, Anthropic, Google, Microsoft and AWS, signed an open letter urging a coordinated surge in cyber defence, citing a closing window before AI-enabled attacks become routine. Notable absences: Meta, Apple, Nvidia, xAI and every Chinese lab. |
|
| The letter calls for treating AI-generated code as a security risk, making defensive AI accessible to hospitals and utilities, and ensuring agentic identities, meaning AI agents acting autonomously, are traceable. There are no funding commitments, deadlines or enforcement mechanisms. |
|
Bottom Line: Tech coalitions are setting de facto security standards for AI-generated code and autonomous agents before regulators do. If your team ships AI-generated code to production, this coalition has named the bar, and your enterprise buyers will start asking whether you meet it.
|
|
|
🇮🇳 India |
Indian IT's AI Productivity Myth |
 Image source: Generated using Google Gemini
|
| India's five largest IT firms together booked $25.6 billion in Q1 FY27, all reporting higher revenue per employee year-on-year. But the long-run CAGR tells a quieter story: TCS managed 0.59% annual growth in this metric since FY19, Infosys 2.47%, and Wipro actually fell. |
|
| Analysts split the driver roughly 60% workforce rationalisation, 40% AI efficiency. Infosys attributes just 8.2% of revenue to AI-led services. The shift from better utilisation to genuine non-linearity, revenue growing without headcount, has not yet arrived. |
|
Bottom Line: Rising revenue per employee is being framed as AI transformation, but analysts put 60% of the gain down to workforce cuts. Indian SaaS and services founders should scrutinise any productivity claim that doesn't separate headcount reduction from genuine output growth.
|
|
Indian labs are putting AI between the test and the doctor |
 Image source: Generated using Google Gemini
|
| Orange Health Labs auto-approves reports and flags abnormal results for doctor review. It has crossed ₹200 crore annualised revenue, is operationally breakeven, and shifted from near-zero preventive testing to 55% of its business, with average order value rising from ₹1,000 to over ₹2,500. |
|
| Curelo layers AI interpretation across 450-plus labs and 4,000 tests, reporting 75,000-plus AI assistant chats per month. Neither company has published accuracy figures or false-negative rates for AI flagging, a critical gap when the cost of an error is a missed diagnosis. |
|
Bottom Line: Auto-approve-with-exception-flagging is the architecture converging across AI agent products. In diagnostics, a missed flag is a missed diagnosis, and neither Orange Health nor Curelo has published false-negative rates or clinical audit data for their AI systems.
|
|
Gnani releases an open-weights Indian language model |
 Image source: Generated using Google Gemini
|
| Gnani AI has released Evon v3.3, a 30-billion-parameter model using mixture-of-experts architecture, meaning only 3.5 billion parameters activate per request, cutting compute costs. It covers 11 Indian languages, runs on a single server node, and is available on Hugging Face under an open Apache 2.0 licence. |
|
| Gnani's core claim: its rebuilt tokeniser needs roughly 20% fewer tokens per Indian-language word than GPT-5's, and less than half what DeepSeek, Llama and Qwen require. Fewer tokens means lower cost per request and more usable context. All benchmark figures are Gnani's own and unverified. |
|
Bottom Line: Tokeniser efficiency, how many tokens a model needs per word, compounds on every API call and determines real running costs more than benchmark rankings do. If Gnani's 20% reduction for Indian-language words holds under independent testing, it reshapes cost assumptions for any Indic AI product.
|
|
|
|
⚡ Mailr Signal |
|
The Gnani tokeniser claim is worth watching closely. Tokenisation efficiency is upstream of every cost and latency calculation in production Indic AI, and if the 20% reduction holds under third-party testing, it shifts the build-vs-buy calculus for any Indian enterprise or government deploying language AI at scale.
— Our take, not a news summary
|
|
|
🛠️ Tool of the Day |
|
Clairva
Converts real-world video into structured behavioral signals for AI
| 🇮🇳 Clairva focuses on India and Global South markets, providing culturally grounded training data for underrepresented languages and behaviors in AI datasets. |
|
| ✅ Founded in 2025 by Sunil Nair, Sabari Raju, Dushyant Verma, and Amit Parashar; raised $500K pre-seed funding led by Venture Catalysts in June 2026. |
|
| 💡 Best for: AI labs and robotics teams needing licensed, culturally-grounded video training data representing real-world human behavior. |
View tool ↗
|
|
|
📱 Social Scan |
 It must have been 5 PM in Claude's world
View post ↗
|
|
|
🧑💻 Founder of the Day |
|
Aastha Rajpal
Ayna
|
| Ayna uses generative AI and diffusion models to create virtual fashion models and studio-quality photoshoots from flat-lay images. Brands customize virtual models across 1,000+ options, 500+ backgrounds, and 100+ templates. Clients include Clovia, WomanLikeU, and Wakefit. |
|
| Indian D2C fashion brands spend Rs 50,000 to Rs 5 lakh per photoshoot with inconsistent results. Ayna replaces expensive, time-consuming traditional photography with AI-driven software, solving a genuine operational constraint at scale for mid-size to large Indian e-commerce brands. |
|
|
|
|
|
पुन्हा भेटूया
That's it from us today. Spotted something we should cover? Know a founder building something interesting? Reply — we're always listening.
Something big is always just around the corner. See you soon.
Were you forwarded this? Join free at mailr.tech
|
|
|
|
You're receiving this because you subscribed to Mailr.
© 2026 Mailr
|
|
|
