Sabitlenmiş Tweet
Deepak Sarda
10.5K posts
Deepak Sarda
@antrix
CTO @ https://t.co/l4js7YgLG2 | Opinions are my own
Singapore Katılım Nisan 2007
253 Takip Edilen239 Takipçiler
Deepak Sarda retweetledi
I strongly believe there are entire companies right now under heavy AI psychosis and its impossible to have rational conversations about it with them. I can't name any specific people because they include personal friends I deeply respect, but I worry about how this plays out.
I lived through the great MTBF vs MTTR (mean-time-between-failure vs. mean-time-to-recovery) reckoning of infrastructure during the transition to cloud and cloud automation. All those arguments are rearing their ugly heads again but now its... the whole software development industry (maybe the whole world, really).
It's frightening, because the psychosis folks operate under an almost absolute "MTTR is all you need" mentality: "its fine to ship bugs because the agents will fix them so quickly and at a scale humans can't do!" We learned in infrastructure that MTTR is great but you can't yeet resilient systems entirely.
The main issue is I don't even know how to bring this up to people I know personally, because bringing this topic up leads to immediately dismissals like "no no, it has full test coverage" or "bug reports are going down" or something, which just don't paint the whole picture.
We already learned this lesson once in infrastructure: you can automate yourself into a very resilient catastrophe machine. Systems can appear healthy by local metrics while globally becoming incomprehensible. Bug reports can go down while latent risk explodes. Test coverage can rise while semantic understanding falls. Changes happens so fast that nobody notices the underlying architecture decaying.
I worry.
English
Why must everything be a binary choice? In this case, it was always about having both. And more! It was even called a test pyramid!
Cindy Sridharan@copyconstruct
end-to-end testing > unit tests, in the vibecoding era. A massive, almost entirely agent-coded refactor passed all unit and pre-merge tests but broke a critical feature. It was only caught due to my own excessive paranoia making me run end-to-end tests before the prod deploy.
English
3 failed attempts at reKYC with @HDFC_Bank. 1 online and 2 over email. Online failed with no reason provided. Email guy stopped responding.
Of the 5 days I'll be in India on my next trip, ~0.5 to 2 days will be spent dealing with this. That's 10%-40% of my trip.
What a mess!
English
Deepak Sarda retweetledi
I'm back in Austin three years after spending my first day at @xai here. On the flight I thought about all the decisions I made during that time - many of which were actually poor. I compiled a selection below.
English
Deepak Sarda retweetledi
Eat all mangoes. Eat every mango you can. The small, juicy, fibrous, nameless ones you pluck off that old tree down the gully. The planted-a-Dussheri-seed-but-never-grafted Dussheri-ish mango from your masi’s garden in Bhopal.
Blushing Sindhuras. Sweet Kesars. Yes eat as many of those beautiful Alphonsos from Devgad packed carefully in cardboard boxes as you can. But also eat Pairi, Neelam, Ratna. And Sindu - a cross between the Ratna & the Alphonso which is slowly gaining more ground because climate change has wreaked havoc on the finicky Alphonso - ask farmers, yields have been down for a few years now, talk to farmers.
Eat with the season.
Eat the early mangoes from the south - Mankurad, Badami, Banganapalli, Imam Pasad in late April-May.
Eat your middle-India mangoes - Kesar, Bombay Green, and also Malgova and Mallika which are late season bloomers from the south, and your Himsagar, Gulabkhaas, beauties from Malda and Murshidabad - try to lay your hands on a Champa, Saranga, or Kohitoor in May-June!
And go both hands in, into piles of Amrapali, Chausa, Malihabadi Dussheris, Langdas in July.
This is just the tip of the mango iceberg, there are so many more loved & delicious varieties - India has near 1,500 varieties of mangoes.
Why would you eat just one? Of course have your favourites but also look at our beautiful biodiversity - please cherish it! Eat widely. Eat greedily. Eat because these mangoes are so delicious. Eat in RESISTANCE TO LOSS, eat like these mangoes might disappear because some of them already are!
Kaushik Subramanian@TheHolyKau
Sheel pls don’t eat tier 2 mangoes There are only two mangoes worth eating in the world, in order: 1a) Alphonso from Devgad (if you can verify). The soil + sea breeze on the slopes gives it a distinct flavour, unmatched 1b) Alphonso from anywhere else in Konkan 2) Imam Pasand The rest are mere fruit. Come to London, I will supply!
English
Deepak Sarda retweetledi
@TheStalwart i think this is the reason they’re algorithmically boosting vagueposting
English
Deepak Sarda retweetledi
Alright Anthropic, OpenAI:
I need APIs that give me usage data. Granular. Per user.
I need this in the same way its provided by every other company for all of time.
What I dont want: "heres a $100,000 line vague line item of token spend".
English
Deepak Sarda retweetledi
This is so sad. I wonder how much money brand consultants make for destroying a very distinctive brand identity and changing it to something entirely undistinctive.
Shobhit Bakliwal@shobhitic
godrej got rid of their old iconic logo
English
This already exists. Kobo + Instapaper sync.
subiii@_subiii_
hear me out - a kindle that's synced with substack + any blogs you follow
English
These pelicans are kind of angry looking!
Left is deepseek-v4-flash, right is deepseek-v4-pro - both generated using OpenRouter via my LLM tool
DeepSeek@deepseek_ai
🚀 DeepSeek-V4 Preview is officially live & open-sourced! Welcome to the era of cost-effective 1M context length. 🔹 DeepSeek-V4-Pro: 1.6T total / 49B active params. Performance rivaling the world's top closed-source models. 🔹 DeepSeek-V4-Flash: 284B total / 13B active params. Your fast, efficient, and economical choice. Try it now at chat.deepseek.com via Expert Mode / Instant Mode. API is updated & available today! 📄 Tech Report: huggingface.co/deepseek-ai/De… 🤗 Open Weights: huggingface.co/collections/de… 1/n
English
I have a vendor who writes completely shitty code and they flourish with making some stupid templated javascript projects for their clients, completely useless. No test cases, no scale, and no scalability.
I can't defame them, but I want to protect future clients to fall for them.
English
Deepak Sarda retweetledi
Mythos is not a bad name for a model but it would be better if Anthropic switched to using famous Claudes. Monet, Debussy etc. The final model that achieves AGI would obviously be Van Damme
English
3.1 Flash TTS is twice the price of 2.5 Flash TTS, and same price as 2.5 Pro TTS.
Demis Hassabis@demishassabis
Our most expressive and steerable TTS model yet! Designed to give builders granular control over AI-generated speech, Gemini 3.1 Flash TTS is really fun to play with! Available in preview today - for devs via the Gemini API & @GoogleAIStudio + for enterprises on Vertex AI
English
At least this limitation is finally acknowledged by them.
Do note that the workaround to split into chunks doesn't work because you very quickly hit the # of requests per day rate limit.
#limitations" target="_blank" rel="nofollow noopener">ai.google.dev/gemini-api/doc…
English
I wonder if anyone is even using these Gemini TTS models. Just try generating anything longer than 5 minutes and you will find the pitch changing and the audio becoming shrill and unusable as the duration progresses.
Even in NotebookLM, you can hear the subtle shift in pacing every 2-3 minutes because they need to concat multiple generated streams to avoid the worst of the duration linked audio degradation.
Oh.. and still no mp3/ogg output!
Demis Hassabis@demishassabis
Our most expressive and steerable TTS model yet! Designed to give builders granular control over AI-generated speech, Gemini 3.1 Flash TTS is really fun to play with! Available in preview today - for devs via the Gemini API & @GoogleAIStudio + for enterprises on Vertex AI
English