AIpreneur; Enable, Educate, & Empower Building Responsible Generative AI Systems; Founder - CloudDon (https://xitter.gq/t.co/rAaOIolrce); MA Sanskrit; (PhD Hindu Studies).
Seattle, WA, USA
Joined December 2009
- Tweets24K
- Following4.3K
- Followers3.2K
- Likes49.5K
Pinned Tweet
My kids are published authors now! Check out their books on Amazon (available at multiple regions):
Bharatanatyam: The Traditional Dance of India (by Srinidhi Sriram): amazon.com/dp/B07XGDB5MM
All About Cars (by Krishna Sriram): amazon.com/dp/B07XFGL1N6
#ProudDad
Watching ‘Modern Family’ dialogues getting muted on Singapore Airlines was not in my bingo card. I should check it out on the likes of Emirates/ Qatar/ Etihad.
How easily could AI replace your job?
While not all jobs can be displaced by AI yet, AI's potential disruption is real. The question I keep hearing is personal: what does this mean for my job?
So I built WillIBeReplaced.ai. Type your job title and get a free job displacement risk score (0–100), based on CloudDon's AI² Index of 738 US occupations. The job displacement risk score for Software Developers is 98.8, indicating a high risk. Surgeons score 18.9, indicating low risk. Where do you land?
Full reports $0.99 until Nov 1. willibereplaced.ai.
@AnthropicAI filed its S-1 today, so I scored my earlier report on its revenue channels (CIIB-006, paywall): 1 right claim, 1 directionally supported, 1 missed. Postulate open.
Cloud marketplaces: 48% of H1 revenue.
Channel fees: ~8¢ per revenue $ (FY25) → ~17¢ (H1 26).
The reach kept climbing. The rent more than doubled.
Check this CloudDon Field Note to learn more about what enterprise buyers should negotiate now, and my recommendations for Anthropic.
clouddon.ai/research/anthrop…
CIIB-006: clouddon.ai/research/ciib-00…
A week since launching Samskritam.ai, and the Library has grown from three texts to eight: the Kumārasaṃbhava, Ṛtusaṃhāra, Kaṭhopaniṣad, Aṣṭāvakragītā, and the whole Rāmāyaṇa, all 18,715 verses of the critical edition. Every word comes with its grammar, checked against Pāṇini’s rules, and now with its meaning too. English translations of the Rāmāyaṇa are going in one kāṇḍa at a time.
The Verse Reader, where you paste any verse, has new tools:
• It suggests the verse’s rasa, showing the words that build it.
• It shows how the words join, with the sūtras behind each sandhi.
• It sets an AI’s scansion beside the rule-based one, so you can see where the model slips.
• If the verse is already in the Library, it compares the AI’s word split with the scholars’.
Try it at samskritam.ai. Corrections are welcome on every page.
Can AI read Sanskrit? Better than I expected!
I built a benchmark graded by Pāṇini's rules on 11 texts the models couldn't recite from memory.
Claude Opus 5.5 scored 89%, Gemini 3.8 Flash 79%, GPT-6 Sol 78%. Their weak spot: meter.
Check out Samskritam.ai to learn more about using AI to learn Sanskrit and how you can contribute.
Sriram Subramanian retweeted
314 years ago, a book of Indian medicine was shipped from the Tamil coast to Germany. It is still there.
Let us hear that story....
The place is Tharangambadi, the Tranquebar of history books, a small Danish trading colony on the Coromandel coast. A German missionary, Johann Ernst Gründler, was posted there by the Danish-Halle Mission.
He sat with an Indian scribe and a local Brahmin, and began collecting, translating and compiling the medical knowledge of the Tamil country.
The result is dated 1712. A handwritten manuscript of 186 pages, titled Medicus Malabaricus. The Malabar Doctor.
It opens with a preface, then a 19 page medical dictionary. Then 132 pages of the real thing. Diseases, their causes, their diagnosis, their treatment. Medicines, their composition, their application.
Gründler sent it home to Halle, the mission's headquarters in Germany. And there it sat, in the Francke Foundations' archive, for 3 centuries.
Only in 2021 did Halle begin a proper project to catalogue and edit it. Workshops at Germany's national science academy followed.
In 2025 it was presented at a Cambridge symposium. The work is still going on. The German archivists call it a source of "outstanding importance for the preservation of India's autochthonous cultural heritage in the history of medicine."
The medicine inside is Tamil, the sister tradition of Ayurveda. Written down with Indian hands in an Indian port, 314 years ago. Unfortunately, we have barely heard of it yet.
Source: Francke Foundations, Halle - "The Medicus Malabaricus of 1712" project. The full manuscript is digitised and readable online in their Digital Collections.
Photo 1: The very 1st leaf of the manuscript: hand-coloured deities, Ganesha on the left, the multi-armed central deity on a lotus. Torn at the fold, but gorgeous.
Photo 2: The manuscript's 1st written page, straight from the archive scan. This is what 314 years looks like.
Photo 3: A diseases chapter opening - Tamil medical terms transcribed into German script (vatham and pitham, the doshas, appear as "Wadum" and "Bittum").
Photo 4: The numbered recipes with Tamil drug names written out one by one: Tschukku (dried ginger), Tippili (long pepper), Nilawembu, Milagu (pepper). Indian medicines, numbered in German, 1712.
Wow!
In the last 2 weeks, one man in India quietly did what AI labs spend billions chasing. And nobody noticed.
Jitesh Prasad, a physics grad from the University of North Bengal and MBA from IIM Lucknow, runs a one man lab called Udgam. Between Oct 10 and 23, he shipped 5 open AI models and a public benchmark.
There is Laghu, a chat model trained from scratch for $146. The full cost ledger is public, failures included: 8 rented GPUs, 13.2 hours, all in < $200.
There is Shruti, a Hindi-English-Hinglish speech to text model in a single 29.5 MB file. It runs offline inside a browser tab.
On the same Hindi test clips it makes fewer errors than OpenAI's Whisper large-v3, a file a 100x its size. On English, he tells you plainly, Whisper wins.
There is Gargi, a 384M reasoning model named after the Vedic philosopher Gargi Vachaknavi. Trained on 17B tokens, about a 1,000x less data than Alibaba's Qwen2.5-0.5B, and within 2 points of it on reasoning tests. On knowledge tests, his own model card admits, it loses badly.
Everything is Apache-2.0 on Hugging Face: weights, code, checkpoints, evaluations, even the $0.45 spent discovering a failed run.
No funding round. No press release. Just one man, thirteen days, and the belief that honest numbers matter more than big ones.
Sources: udgamlabs/blog (link in the comments); Hugging Face model cards udgamlabs/Shruti, udgamlabs/Gargi, udgamlabs/Laghu
Photo 1: Shruti model card WER table - 29.5 MB Shruti beating Whisper large-v3 (3,087 MB) on all three Hindi sets, with Whisper's English wins visible.
Photo 2: Laghu's public cost ledger - $146.26 for the main run, ~$194 all-in, even the $0.03 X search row.
Photo 3: Gargi vs Qwen2.5-0.5B - 384M on ~0.017T tokens vs 494M on ~18T, within 2 points on reasoning, honest MMLU gap in the same rows.
Photo 4: The Hugging Face org page - "Team members 1" next to the full model lineup. The one-man lab in a single screenshot.
Sriram Subramanian retweeted
In the last 2 weeks, one man in India quietly did what AI labs spend billions chasing. And nobody noticed.
Jitesh Prasad, a physics grad from the University of North Bengal and MBA from IIM Lucknow, runs a one man lab called Udgam. Between Oct 10 and 23, he shipped 5 open AI models and a public benchmark.
There is Laghu, a chat model trained from scratch for $146. The full cost ledger is public, failures included: 8 rented GPUs, 13.2 hours, all in < $200.
There is Shruti, a Hindi-English-Hinglish speech to text model in a single 29.5 MB file. It runs offline inside a browser tab.
On the same Hindi test clips it makes fewer errors than OpenAI's Whisper large-v3, a file a 100x its size. On English, he tells you plainly, Whisper wins.
There is Gargi, a 384M reasoning model named after the Vedic philosopher Gargi Vachaknavi. Trained on 17B tokens, about a 1,000x less data than Alibaba's Qwen2.5-0.5B, and within 2 points of it on reasoning tests. On knowledge tests, his own model card admits, it loses badly.
Everything is Apache-2.0 on Hugging Face: weights, code, checkpoints, evaluations, even the $0.45 spent discovering a failed run.
No funding round. No press release. Just one man, thirteen days, and the belief that honest numbers matter more than big ones.
Sources: udgamlabs/blog (link in the comments); Hugging Face model cards udgamlabs/Shruti, udgamlabs/Gargi, udgamlabs/Laghu
Photo 1: Shruti model card WER table - 29.5 MB Shruti beating Whisper large-v3 (3,087 MB) on all three Hindi sets, with Whisper's English wins visible.
Photo 2: Laghu's public cost ledger - $146.26 for the main run, ~$194 all-in, even the $0.03 X search row.
Photo 3: Gargi vs Qwen2.5-0.5B - 384M on ~0.017T tokens vs 494M on ~18T, within 2 points on reasoning, honest MMLU gap in the same rows.
Photo 4: The Hugging Face org page - "Team members 1" next to the full model lineup. The one-man lab in a single screenshot.
Can AI read Sanskrit? Better than I expected!
I built a benchmark graded by Pāṇini's rules on 11 texts the models couldn't recite from memory.
Claude Opus 5.5 scored 89%, Gemini 3.8 Flash 79%, GPT-6 Sol 78%. Their weak spot: meter.
Check out Samskritam.ai to learn more about using AI to learn Sanskrit and how you can contribute.
Sriram Subramanian retweeted
Two classes extremely bad cyber vulnerabilities disclosed in the last 24 hours. The first potentially impacts everyone online. The second impacts nearly everyone running a server.
tl;dr: Everyone should upgrade their OS and Browser as soon as possible. If you run a server you should upgrade PHP (or ensure it’s removed if you’re not using it) as soon as possible.
The first class of vulnerabilities is a set of root TLS certificates that were discovered to have a trivial private key. Literally the number 4. These roots were controlled by #eMudhra. The problem is that with that an attacker could create a certificate and, if they could get in the traffic flow, intercept traffic to any site or service. The solution is to stop trusting certificates issued by these vulnerable roots. Upgrading your browser and OS is the fastest way for most Internet users to do this.
The second class of vulnerabilities doesn’t apply to most consumers directly, but does apply to most anyone running a server. These vulnerabilities have been dubbed #php2shell. PHP is one of the most widely used web programming languages. It’s estimated 70% of websites run on PHP. These vulnerabilities allow an attacker to send a crafted set of queries to a server running PHP and potentially take over the server. From there they could potentially read, modify, delete, or repurpose anything on the server. The solution is to upgrade PHP to the latest version ASAP. Even if you’re not using PHP for your web development, it’s often included by default. So make sure it is either upgraded or fully removed.
@Cloudflare received advanced warning of both of these classes of vulnerabilities. We’ve ensured that where we act as a proxy we are no longer trusting the vulnerable roots. We are also reviewing certificate transparency (CT) logs for any sign the vulnerability was exploited before its disclosure. We have not found any evidence but CT logs are not perfect and a sophisticated attacker who knew about the weakness could have exploited it without a new certificate being logged.
For the PHP vulnerabilities, we have deployed mitigations available for all customers for free. Because of the complexity of the vulnerabilities and number of ways it can be exploited, we also released new WAF capabilities that are unique in the industry to catch the exploit based on the response. These too are available for every Cloudflare customer for free.
Even if you are taking advantage of Cloudflare’s protection, it is important to upgrade your OS, browsers, and PHP installations as soon as possible. We can provide a strong bandaid, but the best protection is to fix the underlying problem.
If you’re interested in learning more, you can read our blogs:
Vulnerable roots: blog.cloudflare.com/emudhra-…
PHP vulnerabilities: blog.cloudflare.com/php-vuln…
Check out the latest episode of #Solvanam This Week in AI (Tamil) to learn more about Dario’s call to pace AI, and a simple intro to open-source/ open-weight and frontier models, with @SolutionArchDJ
youtu.be/MHUXONqShH8
/cc @snapjudge @psankar
Sriram Subramanian retweeted
Gross incompetency from Indian certificate authority eMudhra.
You had one job: generate a 384 bit private keys and instead you used: 4, 56, 66, 115, 155, 156, 230, and 237.
This allows attackers to issue certificates for any websites.
Embarrassing that it’s a family business run by largely computer illiterate people. Only 1 of the 3 have a remotely computer-related background and even they never actually worked a technical job in their life. If you look at the LinkedIn, their tech team is young, from tier 3 education and paid sort of bottom barrel.
Imagine how many attacks you can design with AI given the existence of this genre of gross negligence.
While I’m happy to see @google and @openai are using cricket matches to capture market in India, they need to careful in their claims. While the commentators ‘showed’ creating cartoon style images of @IamSanjuSamson celebrating victory using @GeminiApp, Gemini doesn’t allow that:)
The Tiger is back 🐅
#Jailer2 is in theatres worldwide in a few hours! 51 years in, and Thalaivar still owns the screen. Alappara Kelapparom! 🔥
His whole journey, 1975 to 2026: rajinikanth.ai
@rajinikanth @sunpictures @Nelsondilpkumar @anirudhofficial
Markedly different tone from OpenAI at this DevDay - enabling others to build, sell, and succeed using OpenAI: a welcome move.
Quick thoughts:
1. Dots - underwhelming, late to the market
2. Space - allows Dots/ agents as team members, powerful.
3. Agents API - neat
4. Plugins and Marketplace - cool, have potential
5. Model updates (GPT 6.1 Sol, Ultrafast) - welcome
Misses:
1. No mention of safety, alignment, or security.
2. No such features/ capabilities or new guardrails.
I will be tuning into the @OpenAI DevDay [2026] virtually today. Needless to say alignment and safety will be hot topic, I expect.
I will be tuning into the @OpenAI DevDay [2026] virtually today. Needless to say alignment and safety will be hot topic, I expect.
