Hacker News: daemon_9009

LLM models are not ready for orchestrating many agents

daemon_9009 — Sat, 16 May 2026 20:24:22 +0000

Hi, i have been experimenting with many coding agents distributed in a tree like heirarchy. here is what i found: 1. current LLM models are not trained for orchestrating or managing multiple agents, they prefer doing the work by themselves instead of outsourcing it. 2. you can prompt them to follow the given heirarchy but it is not perfect, main agent will stop the sub agent and perform the work by itself. 3. claude swarms was a feature, but it just does not work seamlessly, i don't want 10 agents to give me a single PR, i want a merged final PR, but you have to ask the main agent to give that, and most of the times, the main agent messes things up while merging all the PRs.

and the only solution is to include these things in the training of the models, current models are individual workers, and not managers. what was your experience with swarms?

Comments URL: https://news.ycombinator.com/item?id=48163477

Points: 1

# Comments: 0

New comment by daemon_9009 in "Ask HN: Are employers getting the returns from AI?"

daemon_9009 — Mon, 04 May 2026 22:18:48 +0000

> >layoff developers for now we are not sure what is gonna happen. Big tech is surely reducing headcount to invest more in AI. but with recent downtime of github/AWS and them not blaming AI but humans, says a lot about that they will go to any extent to prove AI is bringing value enough to fire human engineers.

New comment by daemon_9009 in "Ask HN: Are employers getting the returns from AI?"

daemon_9009 — Mon, 04 May 2026 22:15:30 +0000

if we think, what is accountability? if a human would do a mistake, you would as an employer do two things: either teach him what not to do, or fire him. Same thing can be done with AI agents, if you decide to stay with the agent then teach it what not to do, this might not work 100% but to a certain level for non critical things it would work. Agents should ofc not be given consequential actions like deleting accounts at will. But the point is, once a mistake is done, it is done, be it human or an agent, you gotta teach or fire something. I don't think any other way to solve this problem

Ask HN: Are employers getting the returns from AI?

daemon_9009 — Mon, 04 May 2026 20:45:56 +0000

Hi, most of the companies have now given AI tools to their employees, these include claude code, cursor and github copilot. This is supposed to be the testing period to see how will things turn out by using AI. But now its almost a year, and employers will want to see the return from their investment in AI tools. most devs have still not adopted cloud coding agents for one reason or the other. Layoff number is normal in non-VC funded companies, since they have no pressure to show AI working for employees. But this has to converge somewhere, are returns from AI tools enough to justify the headcount reduction except forcefully showing it to investors?

Comments URL: https://news.ycombinator.com/item?id=48014734

Points: 5

# Comments: 6

Ask HN: How does one review code when most of the code is written by AI?

daemon_9009 — Mon, 09 Mar 2026 19:22:08 +0000

Hi, I spawn multiple cloud agents, but the issue is in reviewing them. even if the model is giving demo artifacts in output, it is not enough to verify all the cases. We have automation test suites, but they are often behind the development. I had tried github copilot review for PRs but it outputs too many nitpicks and false postivies that it becomes irritating to find the real true positive cases! how are you guys handling this surge in code review?

Comments URL: https://news.ycombinator.com/item?id=47314092

Points: 2

# Comments: 0

New comment by daemon_9009 in "[dead]"

daemon_9009 — Sat, 07 Mar 2026 06:15:41 +0000

Hi, guys here are my thoughts on how software development will look in a year. what do you think about this, lets discuss.

Ask HN: What is the current adoption scenario for background coding agents?

daemon_9009 — Fri, 20 Feb 2026 18:30:03 +0000

Hi, background coding agents are here for about half a year now. But only bold and forward teams are using it so far. what is your say? have you adopted background agents in your organization or are most of the people still working with locally running coding agents? One thing i have felt is that organizations are reluctant to give access to these vendors to clone their repos in cloud machines. But what is the overall scenario? are we shifting really to background agents, if yes are organizations and teams trusting third party vendors to clone their repos on cloud machines for the agent to work on? Thanks!

Comments URL: https://news.ycombinator.com/item?id=47091805

Points: 1

# Comments: 1

Ask HN: Why are software developers not using Background coding agents?

daemon_9009 — Wed, 14 Jan 2026 13:19:44 +0000

I have observed that many devs in companies are no t using Background coding agents available in either github copilot or cursor, they prefer IN-IDE agent even though the company is providing them with background agents. I can think of 2 reasons: 1. People don't want to experiment, as background agents seems to be something people cannot control

2. People are doubtful that the agent will be able to complete the task properly.

what do you say?

Comments URL: https://news.ycombinator.com/item?id=46615737

Points: 1

# Comments: 8

New comment by daemon_9009 in "Show HN: What is wrong with the current coding agent workflow"

daemon_9009 — Tue, 13 Jan 2026 11:14:31 +0000

Thanks, we had addressed our pain points, happy to see that you resonate with it!

Show HN: What is wrong with the current coding agent workflow

daemon_9009 — Tue, 13 Jan 2026 10:29:14 +0000

Hi, as the LLM models are getting smarter in coding tasks, we will soon be using agents as co workers, as current tools like github copilot and cursor are not optimized for team collaboration, we began building PhantomX, please give your feedback, if you think we are in the right direction or what should be changed for finding the optimized development workflow which works for both humans and agents.

Comments URL: https://news.ycombinator.com/item?id=46599261

Points: 10

# Comments: 3

New comment by daemon_9009 in "Why aren't smart people happier?"

daemon_9009 — Sun, 09 Nov 2025 09:49:09 +0000

Intelligent people see and understand the world deeper, more than an average person can, They see the world in all its forms, good and bad, understanding too much does not bring happiness but gloom. existence seems to be absurd, the more we understand existence the more one believes that it shouldn't have existed at the first place.

"happiness in intelligent people is the rarest thing I know" is a quote by Hemingway.

Also intelligent people overthink about a particular situation, while its better for survival but it can reduce happniess.

Ask HN: Which coding AI Agents or tools are used in your organization?

daemon_9009 — Thu, 31 Jul 2025 18:44:26 +0000

Hi, Despite using cursor, windsurf or github copilot which are dev tools, have your organization started using background coding AI Agents, where you give them a jira ticket and get a PR? Which tools are you using, and how is the overall experience and which stack have you tried them on?

Comments URL: https://news.ycombinator.com/item?id=44748771

Points: 1

# Comments: 0

New comment by daemon_9009 in "What if we made advertising illegal?"

daemon_9009 — Sun, 13 Apr 2025 05:55:34 +0000

as good as it may sound, but this will never happen. People will never pay for everything, majority don't even buy PC games, they use pirated one's. This is how humans are designed, and this is what which keeps the market afloats.

Ask HN: What else can be automated with AI?

daemon_9009 — Wed, 05 Mar 2025 20:42:45 +0000

Hi, Seeing too many AI products launching, which are automating (up to some extent) tasks. I wonder what else is left which is not in public domain yet?

Comments URL: https://news.ycombinator.com/item?id=43272069

Points: 1

# Comments: 0

Ask HN: What are the Best LLM models for tool use under 7B?

daemon_9009 — Thu, 13 Feb 2025 21:46:38 +0000

I am trying to find a LLM model which is under 7B parameters and can handle json well. I want to use it for tool use locally.

Comments URL: https://news.ycombinator.com/item?id=43041795

Points: 2

# Comments: 0

New comment by daemon_9009 in "Are AI Agents Just Hallucinations in a Digital Dialogue?"

daemon_9009 — Sat, 14 Dec 2024 20:34:08 +0000

I cannot make out anything from this. what are you trying to say? AI agents's fundamental task is to implement stuff, or in technical terms tool calling, computer use etc. they take input from user in natural language, and everything is perfectly defined in the prompt. The LLM is instructed to ask whenever not sure, so bootstrapping from hallucinations only happens when the implementation of AI agent is poor.

Ask HN: Anyone working on the intersection of AI x Gaming?

daemon_9009 — Sat, 14 Dec 2024 19:34:59 +0000

The field of AI x Gaming is quite interesting and i guess gaming is the next industry where we will see the AI boom. From Intelligent NPCs, infinite gameplay to games created using diffusion, as shown by genie2 and other models, are you guys working on any of this? and how do you envision the games next year or so?

Comments URL: https://news.ycombinator.com/item?id=42418949

Points: 2

# Comments: 1

Ask HN: Is working on AI NPCs with LLMs or RL models worth it?

daemon_9009 — Sun, 17 Nov 2024 08:31:12 +0000

Hi guys, i was working on making AI NPCs with LLMs and RL models, to create unique stories for the user, but this is a little far fetched. Is this problem even worth solving? challenges like revenue model, costs of inference and all that. To take an example goodai launched a game https://www.aipeoplegame.com/ , from the first appearances, there is not much we can do in this world. maybe we need a really big open world like GTA to make it a success, but still the question is will people want to play it?

Comments URL: https://news.ycombinator.com/item?id=42162826

Points: 1

# Comments: 0

New comment by daemon_9009 in "Project Sid: Many-agent simulations toward AI civilization"

daemon_9009 — Sun, 10 Nov 2024 18:53:29 +0000

The entire paper demonstrated the results of the simulation or whatever they did. They did not mention how did they achieve this simulation. running 500-1000 LLMs parallely, will take too much computing resources, neither did they prove the claim they made about their parallel architecture. I remeber there was the paper published about an AI town, in which they mentioned clearly how they implemented it. they also released a recording of the simluation along with the real data of the results. If anyone got how they implemented this paper, please tell me.

New comment by daemon_9009 in "Ask HN: What's the most creative 'useless' program you've ever written?"

daemon_9009 — Tue, 29 Oct 2024 18:22:40 +0000

Hi guys, I made a completely multiplayer programming game for C++ and python, started it to learn graphics, multithreading and all. ended up creating a small game engine for my game (included a lot of game maths), learnt multithreading and multiprocessing deeply. the game is originally programmed in c++, so had to do everything from scratch. Later ported it to python using pyBind11, wrote the langauge bindings for python. For the multiplayer part, which took the most of the time, ended up creating my own networking protocol( over UDP) according to the game needs. you can see the project here at my website: https://aiplaygrounds.in