23 | Building with AI | SDE | DM for work & collabs | https://xitter.cf/t.co/dRIhwjmJi6
Joined October 2022
- Tweets6.9K
- Following2.2K
- Followers17.2K
- Likes24.7K
Pinned Tweet
My interview prep used to be 6 tabs:
leetcode for DSA
a youtube playlist for LLD
a random repo for machine coding
a friend for mock interviews (he ghosted)
notion for tracking
panic
here's what replaced each one ↓
now it's one tab. thita.ai
leetcode → 99 DSA patterns that cover all questions asked in FAANG
sliding window, graphs, DP. learn the pattern, then the problems. not 500 random questions.
youtube playlist → LLD + system design
parking lot, BookMyShow, rate limiter. walked through, then you build it in front of a mock AI interviewer , you get instant feedback and can code and submit solution
random repo → machine coding rounds
timed, 90 min, the way Flipkart, Swiggy and Razorpay run them. code has to run. reviewed on structure, not just output.
friend who ghosted → mock interviews with AI
AI interviewer that asks follow-ups, pushes back on vague answers, never reschedules. feedback right after.
notion → progress tracking
it finds your weak spots and sends you back to them. no more re-solving what you already know.
panic → company-wise prep
Amazon OA, Google, Uber, Atlassian. know what they actually ask before you walk in.
~30k people prepping on it.
closed 5 tabs. kept the panic, but it's smaller now.
Most people struggle with DSA because they’re learning algorithms backwards.
They memorize the solution before understanding the problem that made the algorithm necessary.
And then wonder why they go blank when a question changes slightly.
The real unlock is first-principles thinking.
Once you understand why an algorithm works, you can start deriving solutions instead of recalling them.
Patterns stop looking like 500 unrelated tricks. Hard problems become variations of ideas you already understand.
That's how you build actual problem-solving intuition.
@thita_ai Algorithms from First Principles playlist teaches DSA exactly this way.
Wish someone had made this when I was learning to code.
link: youtube.com/watch?v=LfktKt6m…
Asmit retweeted
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
Asmit retweeted
Introducing Kimi K3.1 Pro: our most capable 3T-class open-weight model.
🔹 1M-token context, 2.8T total / 104B active params (MoE), Multimodal.
🔹 Full-project coding: games in Unreal, Unity & Godot.
🔹 Excels at parametric 3D modeling in Blender & CAD.
🔹 More reliable at finance work: Accurate modeling with hallucination down by 65%.
🔹 Matching the strongest proprietary models across coding, agentic and knowledge benchmarks.
Kimi K3.1 Pro and K3.1 Pro Highspeed (2–3× faster) are now live on Kimi.ai, Kimi Work, Kimi Code, and the Kimi API.
Open weights by Nov 4, 2026.
🔗 API: platform.kimi.ai
🔗 Tech blog: kimi.ai/blog/kimi-k3-1-pro
funerals are huge money with zero tech and zero taste
wild bet, love it
i asked chatgpt how to plan a funeral.
it said to call a few funeral homes and ask for price lists. makes sense when only 18% of them post prices online, it had nothing to show me.
every other industry figured out online shopping 20 years ago. this one still runs on phone calls.
so 8 in 10 families call one place and say yes to an $8k decision, sometimes from a hospital hallway.
why are we fine with that?
episode 2 of building an ai-native funeral company in public.
Asmit retweeted
Layoff Announcements in 2026:
1. UPS: 30,000 Employees
2. Oracle: 30,000 Employees
3. Amazon: 16,000 Employees
4. Omnicom: 15,000 Employees
5. Meta: 8,000 Employees
6. Microsoft: 4,800 Employees
7. Dow Chemical: 4,500 Employees
8. Cisco: Fewer Than 4,000 Employees
9. American Express: 3,840 Employees
10. Uber: 3,300 Employees
11. Renault: Up to 2,400 Employees
12. Nike: 1,400 Employees
13. General Motors: 1,300 Employees
14. Disney: 1,000 Employees
15. Citigroup: 1,000 Employees
The tech sector alone has seen 135,000 layoffs so far in 2026 across 330 companies.
Amid rising prices and AI adoption, US corporations are looking reduce expenses and increase efficiency.
Keep watching the labor market.
Asmit retweeted
BREAKING: JPMorgan Chase, Bank of America, Wells Fargo, and Morgan Stanley are scrambling after detecting unauthorized AI agents accessing customer bank accounts, contacting call centers, and reportedly attempting actions WITHOUT instructions from account holders, according to Axios.
