Amal Sinha retweeted
What they want -
OAI: capture light cone of all future value.
Anthropic: rule world because no-one else is worthy.
Google: meet 4th quarter analysts' expectations.
...and that's why I want Google to win.
Folks, I am working on solving this exact problem + many more. In case you have this issue, feel free to reach out.
How do I tell the @Hetzner_Online and @digitalocean CEOs to make new VPS servers:
- have a default firewall that closes all their ports
- by default have Tailscale installed as the only way to access it via SSH
- by default have some web tunnel like Cloudflare Tunnel to serve web traffic safely
That way VPS servers could get a better reputation instantly by being safe by default
It's crazy in 2026 they let you create a VPS that by default is open to the entire internet!
75
Cloudflare launched Clairic which looks like a general-purpose Named-Entity-Recognition tool + Regressor.
Along with Clef/Jev, it feels like we are living through a "general-purpose" version of the 2020 era, when fine-tuning models for specific purposes was the norm.
a brand new model paradigm! introducing 𝚌𝚕𝚊𝚒𝚛𝚒𝚌: a system one model for extracting, computing, estimating, and predicting values from unstructured inputs.
blog.cloudflare.com/clairic-…
119
Anyone who has read Accidental India by @ShankkarAiyar will take this as extremely good news for India. All "great achievements" like food self-sufficiency have largely been triggered by external factors.
India might be the focus of China's new export controls on construction machinery.
India is not called out explicitly, but China is now prohibiting certain machinery to "sensitive countries and regions that have disputes with China" 与中国存在争议的敏感国家和地区.
These include:
- Tunnel-boring machines, which China has appeared to limit to India in the past
- High-altitude concrete spraying machines (above 3,000 m altitude), which might be about military construction in the Himalayas
Amal Sinha retweeted
The recipe is simple:
1. Find repeatable tasks.
2. Use whatever model/size you need to get the task done satisfactorily.
3. Log the runs.
4. Fine-tune the smallest open-weight model that gets you the accuracy you want.
5. Enjoy the benefits of SOTA-level performance at blazing speed on tiny compute.
In that specific case below, I used Parakeet for several months, saved all my audio and transcriptions, and trained the tiny FastConformer on it.
I fine-tuned a small language model after a while, primarily aiming to run a data agent on consumer laptops with no GPU.
Worked a lot faster than I had anticipated. Getting one round of generation done in under 5 seconds.
57
"extremism will never defeat the Abraham Accords" 🤔
Replying to @HamdanMohammed
A hero who stood against terror to prove that extremism will never defeat the Abraham Accords.
@flydubai pilot Smit Machchhar 🇮🇳 is a true model of bravery, perfectly embodying the humanitarian values and spirit of Dubai and the UAE. 🇦🇪🕊️🇮🇱
68
This is the dodgiest and most-desperate thing I’ve seen in the bubble so far. Amazon selling their chips to a group of investors and leasing them back - an off balance sheet arrangement done in plain sight. Really scraping the bottom of the barrel
ft.com/content/97d8d346-519e…
81
Amal Sinha retweeted
$1000-2000 subs are coming over the next few years. If you think that's wild, think again.
Once Anthropic and OpenAI go public, the era of subsidized usage will come to an end.
Secure your compute.
Haha. Maybe its my vision of India?
"Please create an image of what the world would look like if I were in charge, based on my tweets. Go deep, not just the last 10 or whatever is the default. Analyze me."
1
2
88
Amal Sinha retweeted
Replying to @savtrip_dev
Don’t use Figma. I just tell it what I want and iterate from there. Prompts, then screenshots and arrows pointing things out, maybe a picture of a quick sketch on paper, but mostly all just English.
Amal Sinha retweeted
OpenAI and Anthropic oversold AI security breaches to pressure feds into protecting turf: insiders trib.al/Jo6xMk5
Amal Sinha retweeted
Big Short investor Steve Eisman says the AI labs know they have no moats and are manufacturing a crisis to get regulation that hands them a duopoly
"I think this is all nonsense."
[ You think it's all nonsense? ]
"All nonsense. I think that there's something else completely going on here. ... What I think is happening is that token maxing is over. The open weight models are taking big market share. I think these companies are very nervous. They realize that there are no moats around their business whatsoever, and they're trying to manufacture a crisis that will create regulation, and that they think they can then manipulate to create the moats, to create the duopoly that they want."
[ Wow. ]
"That's what I think is going on."
"Honestly, I think this whole Terminator thing is garbage. That's for sure."
LLMs are probabilistic tools being forced to cosplay as deterministic tools for the lazy and dopamine-hungry human mind.
Now we know that AI nearly started a world war this year. It needs regulation but, even more important, the users need education. AI is just a tool - a good slave but a bad master.
timesofindia.indiatimes.com/…
1
107
Anything made using the transformers architecture is a probabilistic model, not a deterministic one. So, such glitches are by design.
EXCLUSIVE: AI actress Tilly Norwood glitches and starts speaking CHINESE during her interview with Piers Morgan and real-life actor Tom Conti...
Watch the full interview at 7pm (BST)👇
📺 youtube.com/piersmorganuncen…
@piersmorgan | @TillyNorwoodX
88
The high one felt with @DSPyOSS has become mainstream.
i think jev is resonating with devs so well b/c it unlocks so many opportunities for composing ai into systems and products rather than ai _becoming_ the product/system
really does feel like it was a missing primitive
1
72




























