🪅 Heap Heap Hurray Ed. #423

Good morning! Let's make this Thursday a success!

#Instagram - More AI slop - this time in search results. To increase the reach of popular posts, Instagram appears to be using AI to generate SEO friendly healines that show up only when you search (on say Google). These are not visible on the post itself - just when you search externally. Which should be ok - except the headlines generated appear to be clickbait and sensational, when they don’t need to be.

#JavaScript - Here’s a throwback to when Netscape Navigator (A precursor to Mozilla Firefox) first announced a new “scripting language” for the internet - Javascript. The internet has never been the same since. (If the link doesn’t load - remove any query parameters in the url - so you end up with dot html at the end of the url and nothing else.).

Google AI Plus Now In India
Google’s AI Plus plan is now availble in India. Seen as a competitor to ChatGPT’s Go Plan, it’s priced at 399 per month - with 199 per month for the first 6 months.

Namma Yatri Partners With Bharat Taxi
Namma Yatri, an Indian ride-hailing startup, has partnered with Bharat Taxi, a government-backed cooperative ride-hailing app, to offer a zero-commission model with fixed fares and driver ownership, competing with Uber and Rapido.

Uber Enters B2B Logistics Space Through Uber Direct And Bangalore Metro Ticketing
Uber India on Wednesday forayed into the B2B logistics space through Uber Direct and launched Bangalore Metro ticketing. Both will be powered by the Open Network for Digital Commerce (ONDC) network. The company says the integration will strengthen interoperability.

Amazon To Invest $35 Billion In India By 2030
Amazon, one of the world's largest e-commerce companies, is planning to invest a massive amount in India to boost its AI capabilities and increase exports. This move comes as the company aims to compete with other major players in the Indian market, including Flipkart and Reliance.

India's First Deep Tech Venture Capital Fund Launched
SINE, IIT Bombay created a 250 crore fund to invest in early-stage startups in AI, advanced computing, and other emerging tech sectors, aiming to accelerate the journey from lab to market.

Truecaller - Family Protection feature launched, enabling users to form trusted family groups for real-time protection against scams and unwanted calls.

Google Photos - Create highlight reels featuring multiple video clips and photos with a single click.

ChatGPT - Added AI Foundations course for practical AI skills training, available in pilot programs with leading employers and public-sector partners, and a separate course for teachers, ChatGPT Foundations for Teachers, available on Coursera.

Reddit - Verified profiles with grey checkmarks for individuals and businesses to confirm their identities.

Adobe Photoshop - Brings industry-leading creative editing capabilities to ChatGPT with conversational interface.

Instagram - Adds AI-powered "Your Algorithm" feature to Reels, allowing users to fine-tune video recommendations based on their interests.

Facebook - Multiple design updates that make it more like Instagram.

Addy Osmani discusses the "AI paradox": while AI tools like Vibe coding increase individual coding speed, shipping is slowed by unscaled human review and validation, with pull request times rising by up to 91%. He stresses that AI excels in low-context, boilerplate tasks (the "70% problem") and that engineers must become "orchestrators" who focus on context engineering, maintaining human control, and evolving code reviews for mentorship to mitigate skill erosion.

The keynote "AWS re:Invent 2025 - Keynote with Dr. Swami Sivasubramanian" focuses on Agentic AI. It introduces tools like the Strand Agent SDK and Amazon Bedrock Agent Core, highlighting features such as episodic memory and reinforcement fine-tuning for model efficiency. The keynote also covers building trustworthy agents using neuro-symbolic AI, reliable automation with Nova Act, and showcases customer success stories from Cox Automotive and Blue Origin.

Suman Debnath from AWS discusses "VoiceVision RAG," a method for integrating visual document intelligence with voice responses. The presentation focuses on using the Pali vision-based retrieval model as a powerful alternative to traditional Retrieval-Augmented Generation (RAG) for processing complex, multimodal data like PDFs containing images and embedded text. He demonstrates how to build an agentic RAG system using the lightweight Strands framework to efficiently retrieve information from document images and deliver the final answer via voice output.

Building A Career As A Tech Speaker: A Step-by-step Guide
To become a successful tech speaker, start by writing a unique talk and presenting it at local user groups, such as those found on Meetup[dot]com. Focus on building your skills and networking with other presenters in the community. In year one, aim to give 4-6 talks, and in year two, target community conferences to get noticed. Follow up with the people you meet and build relationships through social media or email.

Downstream Detector And The Real Cost Of No Upstream Dependencies
Cloudflare's outage in November 2025 exposed a dependency of its own monitoring service, Downdetector, which uses Cloudflare for DNS, CDN, and Bot Protection, highlighting a trade-off between cost, performance, and infrastructure complexity.