Hacker Newsnew | past | comments | ask | show | jobs | submit | ttoinou's commentslogin

If it was so obvious and easy, why didn't we have this a year ago ? Models were mature enough back then to make this work

The high level idea is obvious but doing it is not easy. "Maybe agents should work in teams like humans with different roles and responsibilities and be optimized for those" isn't exactly mind bending. I experimented with it too when LLM coding became a thing.

As usual, the hard part is the actual doing and producing a usable product.


Orchestration definitely wasn't possible a year ago, the only tool that even produced decent results that far back was Aider, it wasn't fully agentic, and it didn't really shine until Gemini 2.5 03-25.

The truth is that people are doing experiments on most of this stuff, and a lot of them are even writing about it, but most of the time you don't see that writing (or the projects that get made) unless someone with an audience already (like Steve Yegge) makes it.


Roo Code in VSCode was working fine a year ago, even back in November 2024 with Sonnet 3.5 or 3.7

Because gathering training data and doing post-training takes time. I agree with OP that this is the obvious next step given context length limitations. Humans work the same way in organizations, you have different people specializing in different things because everyone has a limited "context length".

Because they are not good engineers [1]

Also, because they are stuck in a language and an ecosystem that cannot reliably build supervisors, hierarchies of processes etc. You need Erlang/Elixir for that. Or similar implementations like Akka that they mention.

[1] Yes, they claim their AI-written slop in Claude Code is "a tiny game engine" that takes 16ms to output a couple of hundred of characters on screen: https://x.com/trq212/status/2014051501786931427


Do you think 65 inch would still be OK for a monitor usage ? I've been pondering about doing that for years but 65 inch is often easier to find for me in Europe

Monitor - I prefer higher refresh rate. The other part is that you should have the entire picture in your eyesight w/o moving the neck.

Personally, I'd consider that large of a screen, a bad working area.


>The other part is that you should have the entire picture in your eyesight w/o moving the neck.

I agree, but that is still the case here. The difference is that the "full picture" no longer occupies the whole monitor.

The amount of windows and content you arrange on a normal office monitor is about one third of my available space. I can arrange my windows in that space and not have to move my neck.

But at a glance, I can also see the contents of 4 other code files in my project that are also visible, as well as my notes, the documentation, the team chat.

Or if I want, I can also see twice the amount of code in any file by having my editor take up the full height of the center third of this monitor.

Basically the monitor goes from being the "full picture" to a canvas where you are free to create any collection of a "full picture" you want, and you can have the leftover building blocks visible on the sides, optionally.

I am sure that if you let all knowledge workers in the world test this setup for a day, a vast majority of them would want to keep it. But since even 8k tv's are going away now, most will never know.

Curved gaming monitors costing more than my TV are being deployed everywhere lately, for productivity work. Most people are used to 27" or 24" low-res monitors and they are getting an upgrade, but i's not a very good one.

Had the panels from 8k tv's been used in monitors and marketed to corporations it would have been so much better!

Perfect for open offices too - no need for desk dividers if everyone is behind a huge screen! ;-)


I have been using ultra wide monitor for years.

>Perfect for open offices too - no need for desk dividers if everyone is behind a huge screen! ;-)

Folks have been using 2-3 monitors for quite some time with the same purpose, but I also don't get the idea of the dividers.


Nice thanks. What OS are you using and what tools to organize windows ?

I have several computers attached to the screen and run Windows, Linux and MacOS.

Of these I think Windows has the best solution with a utility called Fancyzones (part of a collection called Powertoys). I can define areas of any size and shape and easily move windows around so they occupy the defined spaces.

In Linux I have used KDE, which has at least one similar mechanism which works fine too, but was more trickly to set up and configure.

Lately I have also been running hyprland more often and the setup actually works well with a tiling WM. You can tile many times before the windows get too small to be useable!

One thing worth mentioning is that it's kinda tricky to get 8k/60hz working under linux since the TV only does HDMI. I tried several 8k DP to HDMI adapters before I found one that worked. Windows and Mac work fine with a normal 8k HDMI cable.

On the mac I just use manual window management, since most of what I do on the mac tends to just use one main window anyway (garage band, lightroom, photoshop and such).


With one monitor you have to see the whole screen but with multi monitor setups extra screens may be more like digital photo frames.

If you're sitting about a meter away, that's probably fine. Most desks aren't deep enough, and any closer would be a real problem for head movement.

But if you have it wall mounted at eye level or a deep desk you're likely okay.


I don't think it will work very well as a "normal" monitor (meaning placed at normal monitor distance on your desk).

My 55 is borderline too big already, and the main issue is actually the height. Tilting your head or rolling your eyes back to see the top gets noticeably uncomfortable pretty quickly.

I made a special mount so the lower edge basically rests on the desk surface which basically solved that issue, but I don't think I could have made it work if it was any bigger.

Also at 65 the pixel density is much lower, so you'd probably want it mounted further away. But if you do, the monitor will cover the same FOV as a smaller monitor anyway.

My dream is that someone starts making 8K 50" monitors with displayport inputs (HDMI is a mess) and sells them for the same price as these tv's used to cost!


Yet this was released under their term and not previous presidential terms

> previous presidential terms

Term, not plural. There was one (1) interceding administration following Epstein's death.

Trump promised that the Epstein files would be released if he was reelected, and then withheld files. Congress passed a bill remediating this, hence the newer tranche of files: https://en.wikipedia.org/wiki/Epstein_Files_Transparency_Act


Im not too deep into USA politics and have very very bad memory so i dont remember how it went down. The Wikipedia article you linked says it was signed by trump.

>and then withheld files.

So did he sign that willingly in the end? Did he have to sign it? Did he cave because he said publicly he would?


A super majority in Congress voted for the files to be the released, which is enough to override a veto, so he had to sign it to save face.

thx for the info!!

So how do you get this ?

Let's Encrypt can issue wildcard certs too

What happens if we use Claude Pro or Max plans on them ? It’ll always be a different IP connecting and we might get banned from Anthropic as they think we’re different users

Why limit the lifetime on 30 mins ?


We'll increase the lifetime in the next weeks - just some tech internally that needs to be adjusted first.

For what it's worth, I do this from about 50 different IPs and have had no issues. I think their heuristics are more about confirming "a human is driving this" and rejecting "this is something abusing tokens for API access".

All the time with the same computer ? Maybe it is looking at others metadata, for example local MAC addresses

All the time with a bunch of different sandboxes.

What's the use case for this? Trying to get raw API access through a monthly plan? Or something else?

Simply using your subscription in a sandbox ?

I can run nightmedia/qwen3-next-80b-a3b-instruct-mlx at 60-74 tps using LM Studio. What did you try ? What benefit do you get from KV Caching ?

KV caching means that when you have 10k prompt, all follow up questions return immediately - this is standard with all inference engines.

Now if you are not happy with the last answer, you maybe want to simply regenerate it or change your last question - this is branching of the conversation. Llama.cpp is capable of re-using the KV cache up to that point while MLX does not (I am using MLX server from MLX community project). I haven't tried with LMStudio. Maybe worth a try, thanks for the heads-up.


Cerebras already has GLM 4.7 in the code plans

Yep. But this is like 10x faster; 3B active parameters.

Cerebras is already 200-800 tps, do you need even faster ?

Yes! I don't try to read agent tokens as they are generated, so if code generation decreases from 1 minute to 6 seconds, I'll be delighted. I'll even accept 10s -> 1s speedups. Considering how often I've seen agents spin wheels with different approaches, faster is always better, until models can 1-shot solutions without the repeated "No, wait..." / "Actually..." thinking loops

> until models can 1-shot solutions without the repeated "No, wait..." / "Actually..." thinking loops

That would imply they'd have to be actually smarter than humans, not just faster and be able to scale infinitely. IMHO that's still very far away..


Does it look like functional programming anymore ?

Looks pretty ML:ish to me, even in a segment like this:

   let parse_int64 (local_ buf) (sp : span) : int64# =
     let mutable acc : int64# = #0L in
     let mutable i = 0 in
     let mutable valid = true in
     while valid && i < I16.to_int sp.#len do
       let c = Bytes.get buf (I16.to_int sp.#off + i) in
       match c with
       | '0' .. '9' ->
         acc <- I64.add (I64.mul acc #10L) (I64.of_int (Char.code c - 48));
         i <- i + 1
       | _ -> valid <- false
     done;
     acc

Yes - high-performance Haskell code looks similar. There isn't much to be said there - it's a little less clean-looking because FP optimizes for the most useful scenario and trying to do highly advanced stuff like that will be more verbose. This is in contrast to OOP where everything is verbose, and sometimes high-perf stuff that falls into the shape of globals + mutation + goto looks very succinct.

Looks like 100% idiomatic normal OCaml to me.

Technically you are right but too much mutation for my tastes and probably many other ocaml developers.

(author here) The mutation is only for performance critical code. I'm first trying to match C/Rust performance in my code, and then transform it to more idiomatic functional code (which flambda2 in OxCaml can optimise).

It's too difficult right now to directly jump to the functional version since I don't understand the flambda2 compiler well enough to predict whta optimisations will work! OxCaml is stabilising more this year so that should get easier in time.


I think there are more succinct snippets in here and some this more verbose exposition is for pedagogical purposes. I am not a fan of ocaml because tacking on the object syntax made SML more verbose (ugly imo). Looks like 0xcaml continued trend.

OxCaml is OCaml, it is only a set of language extensions that Jane Street expects eventually being able to upstream, depending on the experience.

Yes much like the Object extensions added to Caml.

Depends on what one means as FP.

When I learnt FP, the choice was between Lisp, Scheme, Miranda, Caml Light and Standard ML, depending on the assignment.

Nowadays some folks consider FP === Haskell.


Even F# looks like good FP to me. But yes I expect something short in FP to clearly see the structure of the program, side effects, flow and data

Yeah but you have to watch Man with a Movie Camera with a good soundtrack like The Cinematic Orchestra one. Then it becomes the best non verbal movie ever.

I'll add it to my list!

Brain drain to USA, lack of good VC / funding, Patents war / patents controlled by the USA, industry espionage, EU mentalities that don't favor innovation in the same direction than the USA, too diverse markets to serve, distributed controls / governments / decisions. The list can go on and on

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: