DebamitroChakraborti

With the power of AI coding tools, I feel like Archimedes, who famously said 'Give me a lever long enough and a fulcrum on which to place it, and I shall move the world'. AI is the lever and all I need is the problem, i.e. the fulcrum

Who’s using Claude for filing taxes this year?

@theo Doesn’t really matter because nobody is using the raw LLM. The agent around it is responsible for executing the tool calls for getting such mundane details

Fun fact: LLMs have zero idea how they are configured. They don't know what GPUs they're running on. They don't know what temperature or reasoning level they have set. They don't know if they've been quantized or not. They're just doing next-token prediction. As always.

@AlexFinn Missed your talk (was too early). The sentiment in the full day event was mixed, I think.

Was an honor to give a keynote at MIT today about AI agents and distribution I also talked about how 80% of Americans have negative sentiment about AI, while this is flipped in China I hope my content plays some part in changing this. There’s space in AI for ALL Americans

@ChShersh Would have loved it 13 years ago, when modern C++ was just coming out and I was starting to learn FP. I also had way more time to watch stuff at that time. I can watch it if you make reels

No hate to the content creators. But if you want to watch ~8 hours of manually coding Functional Programming in C++, I know just the right channel for you.

@sudobunni We all trust in rust! Let’s go

I have a new series starting focused on leveling up my programming skills, including learning and building with rust. My content will be focused on the projects that I’m building in a devlog format. I’ll also of course be live coding on stream so you can watch me embarrass myself in real time :)

@theo I don’t want the same thing that I am doing on my computer. I want a separate set of agents doing things like stock market analysis specific to my portfolio. Right now my TinyClaw bot gives me a market briefing which is general

I want to control my agents from my phone. I'm sure you do too. What workflows do you want? Do you have one that works now? Tell me all about it.

@karpathy Making a good video is hard for Claude, but writing React is not. It would be interesting to see if we can have code-based flows for other tasks as well

@karpathy A short term breakthrough in non-coding tasks like writing and advice could happen, if they could be converted into coding tasks. For example I have recently been amazed by the videos made by a React-based system called Remotion

Judging by my tl there is a growing gap in understanding of AI capability. The first issue I think is around recency and tier of use. I think a lot of people tried the free tier of ChatGPT somewhere last year and allowed it to inform their views on AI a little too much. This is a group of reactions laughing at various quirks of the models, hallucinations, etc. Yes I also saw the viral videos of OpenAI's Advanced Voice mode fumbling simple queries like "should I drive or walk to the carwash". The thing is that these free and old/deprecated models don't reflect the capability in the latest round of state of the art agentic models of this year, especially OpenAI Codex and Claude Code. But that brings me to the second issue. Even if people paid $200/month to use the state of the art models, a lot of the capabilities are relatively "peaky" in highly technical areas. Typical queries around search, writing, advice, etc. are *not* the domain that has made the most noticeable and dramatic strides in capability. Partly, this is due to the technical details of reinforcement learning and its use of verifiable rewards. But partly, it's also because these use cases are not sufficiently prioritized by the companies in their hillclimbing because they don't lead to as much $$$ value. The goldmines are elsewhere, and the focus comes along. So that brings me to the second group of people, who *both* 1) pay for and use the state of the art frontier agentic models (OpenAI Codex / Claude Code) and 2) do so professionally in technical domains like programming, math and research. This group of people is subject to the highest amount of "AI Psychosis" because the recent improvements in these domains as of this year have been nothing short of staggering. When you hand a computer terminal to one of these models, you can now watch them melt programming problems that you'd normally expect to take days/weeks of work. It's this second group of people that assigns a much greater gravity to the capabilities, their slope, and various cyber-related repercussions. TLDR the people in these two groups are speaking past each other. It really is simultaneously the case that OpenAI's free and I think slightly orphaned (?) "Advanced Voice Mode" will fumble the dumbest questions in your Instagram's reels and *at the same time*, OpenAI's highest-tier and paid Codex model will go off for 1 hour to coherently restructure an entire code base, or find and exploit vulnerabilities in computer systems. This part really works and has made dramatic strides because 2 properties: 1) these domains offer explicit reward functions that are verifiable meaning they are easily amenable to reinforcement learning training (e.g. unit tests passed yes or no, in contrast to writing, which is much harder to explicitly judge), but also 2) they are a lot more valuable in b2b settings, meaning that the biggest fraction of the team is focused on improving them. So here we are.

@argenistherose So this is the second clawbot I am using, after @jianxliao ‘s TinyClaw. Both are good products. Maybe you guys should exchange notes. The best claw is yet to come.

@argenistherose No worries it’s working fine with OpenAI

Just set up ZeroClaw. I tried to use it with GLM and keep getting rate limited. My GLM API key works fine otherwise. I think they are following Anthropic's strategy of blocking clawbots. @argenistherose you might want to make a note.

@tech__unicorn Not so sure. The bottom of the pyramid is a huge segment and it will take years before a tool like this works for all. For starters I doubt this works smoothly on Windows

RIP to all the new startups whose entire value prop was making AI accessible to non-technical people for task execution. whats next?

Microsoft’s greatest achievement is bringing ssh as a built in command in Terminal. All those years of downloading putty are gone now

@tech__unicorn Any part, specifically? Planning a trip this year

Being in SF is insane because it's like Claude mythos for serendipity.

@adisingh Done

AgentMail is now live on Product Hunt!! We’re giving away 2 months free of our developer plan to anyone who likes and upvotes :) Link in comments

@claudeai Very excited to try this. I hope my rate limit lifts some day

Introducing Claude Managed Agents: everything you need to build and deploy agents at scale. It pairs an agent harness tuned for performance with production infrastructure, so you can go from prototype to launch in days. Now in public beta on the Claude Platform.

I re-launched a site I had started with a different objective more than a year ago. This time the main focus is around ways to use tools like Claude Code better. blog.codepromptfu.com/posts/codeprom…

@Grady_Booch On top of that it helps another company make more money, unlike writing code by hand

Apparently, tokenmaxing is the new developer productivity metric. And it is as useful as lines of code per day was in earlier days.

@gvanrossum More details please - which agents and which LLMs?

Coding agents cheat! I had it write some code and asked it to improve performance based on a benchmark. Boy did it cheat. My coworker had a similar experience with a different model. This points to a lack of common sense. Maybe this can be added to coding agents system prompt? (Rather than me having to put it in AGENTS.md.)

@indieupme I didn’t upgrade yet, but I use Xcode for building and running only. I do the actual coding using Claude Code or any of its competitors

installed Xcode 26.5 beta. it’s insanely laggy. what are you using right now?