
Building intelligence for behavioural health @empaithy | host @ https://teapaw.org/t.co/BN4cFv4K9K
India
Joined June 2009
- Tweets3.2K
- Following819
- Followers1.3K
- Likes18.6K
Harshdeep Anand retweeted
GTAV is a masterpiece. Probably closest we will ever see a handmade game get towards the potential of its own genre.
The reason why you see all the negativity is because negativity sounds smart, and therefore gets engagement. Just play this singular game. You will see.
To Celebrate GTA 6 Release, Indian Company Declares Holiday For Employees, Interns And Freelancers On Nov 19 ndtv.com/offbeat/to-celebrat…
71
Harshdeep Anand retweeted
the illusion of human exceptionalism crumbles fast when you see how much of our intelligence is just pattern recognition & predictive modeling. ai exposes how much of what we call “creativity” is remixing, how much “insight” is regurgitation, & how much “expertise” is just memorized heuristics.
Harshdeep Anand retweeted
Attacks will accelerate from here on - We are there: Six months ago I said Mythos like capability will be widely available. It's happening. A critical vulnerability, a social engineering attack that looks innocuous, or daisy chaining publicly available vulnerabilities and conducting a sniper strike is all being witnessed in the wild. Incident responders have never been busier.
Some key points:
- attacks are not becoming more sophisticated, it's the same attacks, faster and by a lot more people.
- attack path discovery which took trial and error and multiple attempts is now trivial, open source LLMs are here to help
- you are as secure as your supply chain, insecure down or upstream players who have access to your infrastructure could be the next attack path, marketplaces and supply chain is an easier vector
The increase in activity is a result of more people becoming better at hacking due to the ease of being able to attack. Had not contemplated that in my calculus.
Please:
1. Test your infrastructure for accessible vulnerabilities and misconfigurations
2. Fix your code and demand your vendors do.
3. Old infra, or insecure infra at the perimeter needs to be updated first
4. New AI tools, especially coding need a security framework. Your developers will do more with code, new skills, and more automation. Don't forget the coding models have trained on human code.
Stay safe this Halloween :)
Harshdeep Anand retweeted
A ‘hello world’ milestone for Project Suncatcher 🚀
After launching our prototype satellite a few weeks ago, we’ve now successfully sent a command up to orbit, booted up the TPU, ran the prompt, and received our first complete response back on earth - a haiku!
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
What an experience. Recommend this one.
1
39
Harshdeep Anand retweeted
The dire wolf pups just had their biggest check-up yet. 🐺 CT scans, heart scans, X-rays, bloodwork, even whole-genome sequencing.
The verdict? Happy and healthy from snoot to tail.
But how big are they? Scroll for the official weigh-in. ⬇️
(1/10)
Google opened Gemini 4 Argon to everyone yesterday.
The model watched a video, built its own simulation, and helped teach a humanoid to sink a run of cornhole throws. Three pointers next?
1
77
AI is very good at sounding like it understands you. That's precisely what makes it risky in mental health.
Feeling heard and being helped are not the same thing, and a model optimised for the first can quietly work against the second.
5
160
Your therapy journal: Empaithy
empaithy.com/journaling


















