If you manage AI call centers/ sell AI services, read this till the end. Vonage is the world's second-largest telephony vendor. It’s what powers your SMS OTPs, calls, and more. It's like Twilio, but better. Four days ago, Vonage SMS APIs were down for 36 hours. For 36 hours, billion-dollar companies weren't able to send authentication or communication SMSs. This is not the first time Vonage or Twilio has been down. Here is exactly what happened: Vonage's services rely on NorthC's colocation services for carrier and supplier integration. All DIDs (virtual numbers) were dependent on NorthC for carrier integration. Think of it like this: the data center served as the router for digital numbers to interact with physical network infrastructure ie Vodafone virtual numbers interact with Vodafon physical cell towers. On the morning of Thursday, May 7, 2026, a fire broke out in a technical room at the rear of the facility in Almere, Netherlands. The emergency power systems and diesel generators caught fire, causing a complete power outage. When the Almere site went dark, the physical "pipes" connecting Vonage’s platform to local Middle Eastern and European carriers were severed. This should have never happened: The actual virtual numbers were not down. The carriers themselves were not down. The services to send SMS and the switching were not down. It was Vonage that was the single point of failure. Why would anyone with a meaningful business and you are this: Pay a 30% tax to Vonage or Twilio? Run into errors that last 36 hours? When, in reality, there is no real-world issue at all? Every contact center should own their own PBX. This way it is cheaper AND better managed. This also ensures 100% uptime and backups, and you get to be in control of where your numbers are hosted and how the data is processed Companies like Vonage and Twilio are graduating into AI companies, hiding it from you that they are the reason that your entire contact center might stop working. Over the last couple of weeks, I have helped two million-dollar-plus businesses migrate from Twilio to self-managed phone booth services. And we can do that for you as well. Text me now

Fable (noun): a short story that teaches a lesson. Nominative Determinism.

Always check your DMs.

The kind of generational idea I pray maa saraswati hits me with one day.

Malcom Gladwell advocates an idea of behaviour cascades where one economic event cascades to multiple participants resulting in the pie becoming bigger. . The idea that "If South Korea Taiwanand Singapore become the new messiahs of a changing world order, what will India do" often bothers rationalists pessimists and idiots (me) alike. Here is my going theory- the world has only ever seen the success of one generic economy. (Rem Koolhans calls China a generic economy because their paddy fields make bread and semicon with the same spirit). The world however hasn't seen a "non convex economy" evolve to its critical maximus. My case for India is that (I will come back to how this relates to the tagged post in a bit) . - India moved from convexity (wealth concentrated to government owned funds) to relative non convexity (few more wealth builders). - But we are now in an Arrow Debreau instability i.e supply demand and consumption aren't even in line with each other. Everything fucking sucks and everyone is irrational. India while it consumes everything it produces, also doesn't produce what it doesn't consume. It's supply chains are convex (concentrated to few ports and few people) and there's very few people who know what is happening . The good news is that despite all that, we seem to have decided that we are in fact a generic economy. We have also decided to sell human resources for foreign capital. . This is a good step so long as we create enough supply chain surplusses. . Which positions us as- 1) A net consumer for a few decades 2) A net producer of human labour for a few decades. A true blue ocean equilibrium . This is also why I think the pie becoming large helps us. As the number of "first world countries" increases, the number of "second world countries" will grow too. . This will drive up consumption for us, production for others and more forex for us. And everyone else who sticks around . This will create many more firsr world countries. India included. Until then, we hope our position as the handshake between producers and consumers (of human resources) remains consistent. . More first world is imminent. It's just not paced for optimism. For anyone. . Forex and capital markets are perception indicators not mirrors. There will be more fringe first world than everyone ever anticipated.

I studied every employee of OpenDoor India who was fired recently and saw something very interesting 1. 130 out of the 200 employees are "Operations Associates" or customer support. Their tasks were to call the customers and verify their listings. source: LinkedIn bio of three people who work in the same space. It sort of makes sense that this could be automated by AI. Opendoor is also hiring an AI Operations Associate in North America, where it explicitly states that it wants to use AI for operations and customer support. 2. 25 people were in accounting support / accountants- their job was to handle payroll and general bookkeeping. This seems bloated for an org like Opendoor. They are now hiring new accountants in North America who can use AI. 3. There only seem to be 25 Engineers in India- large amount of which are Data Engineers, data scientists (8-10 ) and their DE stack seems.... bloated. I found this LinkedIn post where they discuss building migration tools for their CDC infrastructure. DE cuts are probably cost cuts 4. 10 engineers are QA engineers, 10 are analysts 5. The culture of hiring to fire seems prevalent based on their Glassdoor reviews from up to two years ago. The attrition rate was really high in the org anyway. 6. A large amount of operations and customer support people seem to be from the same university in the same city. 7. Overall, this is largely a nothing burger. Most people were leaving anyway. The people who have been fired are not engineers or high-paid employees, from my assumption.

Fable is a very special model. It feels very "reliable". I assigned it the task to develop a new hierarchical tree embedder that uses leaf nodes as embeddings and supports new data too without retraining. (Deterministic embedding oblivious to new dats) Because it's a tree it should contain rule delegation and chunk delegation within itself. Every embedding can be distilled to a function. Each node is a vector card. One shotted it. Did better than I could. Core ideas ---------- 1. Chunking is learned from the corpus: frequent character n-grams form the chunk vocabulary, with guaranteed single-character fallback. Any string, seen or unseen, decomposes into existing leaves. The leaf set is closed, so the tree never needs new leaves. 2. The tree is built once via recursive bisecting 2-means over PPMI co-occurrence features of the chunks. Leaves are chunks. Every node (internal and leaf) holds a compact weight "card" (a small vector). 3. The embedding of a chunk is the normalized sum of the cards on its root-to-leaf path. Words, sentences, and documents compose from chunk embeddings (frequency-weighted mean). The leaf nodes in totality therefore define the entire embedding space. 4. Training is skip-gram with negative sampling over chunk streams. Gradients update node cards only. Incremental data updates the same cards online. Tree topology is never touched, which is verified with a structure hash before and after the update. Good stuff.

This dude scammed my client off a couple of GitHub copilot licenses by giving everyone free trials instead of a $20 license. It only came to picture because I was helping one of their employees integrate an automation using copilot 😭 (Client is a well known MNC)

No house party has the aura of my parties. 2 Arab dudes, some eurotrash, me and my brother Aviraj (who's now a famous dating coach) and idk what the fuck I was doing. Our home owners suspected we had a drug and prostitution racket and got us out by force. 2021-22. Golden

After using Nothing Phone (3) I want to permanently abandon my iPhone Except for my airtags, theres really nothing iOS really offers. The speakers are so good on Nothing, the screen is snappier, WisprFlow works better, the Essential button is beautiful, the haptics arecrazy too