←
Amanpreet Singh675 posts
Amanpreet Singh avatar

Amanpreet Singh

@apsdehal

something new

◷ Joined January 2010▧ 42 Photos and videos
Connor Shorten avatar
⟳ Amanpreet Singh retweeted
Connor Shorten@CShorten30·Nov 16

Hey everyone! I am super excited to share our new research report! 🎉 Scaling Reranked Retrieval 🚀 Thread with more details! 🧵 (1/17)

Post attachment
Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 3

wow

We asked GPT-6 Astra to edit Qwen3.5-0.8B weights to improve math. It directly changed only 0.016% of weights, fixing 0.8B's reasoning loops & improving math up to 6x and even instruction following 3x, without gradients or training. We call this Weight-Space Programming: frontier LLMs can now natively read/write neural network weights! 🧵 Blog: rikiyat.github.io/blog/weigh…

Post attachment
Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

Building for knowledge work? it's an important question for your platform: when should the user talk to an agent, and when should they be able to interact directly with what it made? We have a strong thesis here and have built a solution around it. Would love to chat. DM! 9/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

Finally, there is another benefit here: implicit feedback. Through user edits you can see what they wanted to change and why agent got it wrong. These become an implicit signal you can learn from, without making the person explain the same correction again. 8/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

We’re already seeing signs of this. OpenAI’s Intelligent UI brings interactive elements and tools directly into ChatGPT. ChatGPT apps bring well-known application interfaces into chat. OpenUI gives any developer a framework for easily generating interfaces from LLM outputs 7/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

As models improve, they’ll get better at this. But, making a specific change to documents is still much easier in the interfaces built for them. These have been refined over years for human use. We'll see more of these inside the chat. 6/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

This is even more important in the knowledge work. Say you are working with Excel and having an agent fill out a financial model. Now you want to change a single cell. Do you trust it to change only that cell? Recalculate everything correctly and leave everything else alone? 5/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

This loop continues leaving you frustrated If you've used LLMs a lot, you know this feeling! Enter UI components embedded directly in the chat Grokbot solves this by putting a basic email editor in chat itself Small thing but fixes a real limitation of chat and iterations 4/n

Post attachment
Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

For example, take email automation with Grokbot. It can read emails and write a draft. Often you want to only change 1 sentence, who’s CC’d or BCC’d, or fix the subject. You ask the agent to make the change. But, did it change anything else? Now, you've to check again. 3/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

The frontier labs have now realized that the chat is great for a lot of things, but doing something very specific requires two things: (i) describe their ask precisely (ii) the model to follow that ask accurately. Both are hard to guarantee. 2/n

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

A year ago, everybody was pushing chat as a universal interface in their applications. Every system of record, every SaaS application. But, as agents start to do actual work, the reverse is happening and the UI is now showing up embedded in the chat. 1/n

GPT-6 and Intelligent UI, now rolling out in ChatGPT for everyone. Intelligent UI in ChatGPT delivers fast, interactive answers that make everyday questions more visual, complex topics easier to grasp, and interactive tools for your task available on the spot.

1:10
Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Nov 2

Good article. It gives you a glimpse of how the interaction with agents will look in the future as they start doing more work. Super Agents are emerging as a new interaction layer across the board: projects in Claude Code and Cursor also build on the same idea. As we trust agents more, the bottleneck shifts from doing the work to reviewing the work. Over time, the super agent needs to be proactive while only highlighting and bubbling up new exceptions, questions, and edge cases to you.

DeepSeek Harness @DeepSeekHarness Nov 1 Article Exploring Super-agent Design: Towards a More Autonomous Agent System The original article is in Chinese. This is a machine translation. This article discusses architectural ideas. It is not an introduction to existing features of DeepSeek Harness. 1. Super-agent Over

Gurnoor Singh avatar
⟳ Amanpreet Singh retweeted
Gurnoor Singh@gurnoorsingh123·Oct 29

Had a great time speaking at Bench Summit today about the challenges of evaluating AI agents in real-world environments! Lots of interesting conversations with people building and evaluating agents at the frontier. Thanks to @kaggle, @GoogleDeepMind, and @zapier for putting together a great event!

Post attachmentPost attachment
Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Oct 21

Eliciting the tribal knowledge in people's heads that was never written down anywhere is an art and the biggest barrier to the diffusion of AI in enterprises.

"company brain" startups that construct information from internal messages and documents when they realize 90% of information only exist in peoples heads: 😭😭😵‍💫😵‍💫

Amanpreet Singh avatar
Amanpreet Singh@apsdehal·Oct 10

This is good timely note on trustworthiness and safety of enterprise agents. As models and harnesses commoditize, we need to rethink how we can teach agents efficiently and verify their work at scale. It’s not just what an agent can access or do, but who can teach it and who can update its behavior. Our access controls and interfaces also need to evolve with this. The people teaching and updating the agent might not be the same ones verifying its work and signing off on it. The system needs to enforce that separation allow verifying the work efficiently. This is a layer that should work across models, learn from demonstrations and corrections, and provide decision trails as it evolves: what changed, why, who taught it, who verified it. It should also capture how the agent made decisions and on what basis when it did the work. Over time, through learning, it should reduce the verification burden as well. If you can verify it, you can solve it through AI. If you can verify it at scale, you can automate it through AI. Otherwise, every update to the agent becomes another thing we have to trust without being able to evaluate or verify.

Satya Nadella @satyanadella Oct 10 Models as Insider Risks in the Super Intelligence Era As traditional software systems were being deployed across the economy over the last few decades, we had the tools and capability to trace behaviors to a specific code path. That same kind of