Erdal

AI Engineer | 🇺🇸🇫🇷🇹🇷

[Erdal @ErdalToprak]

For months OSS 120B was totally awesome

[Erdal @ErdalToprak]

Extremely shortsighted answer

In many ways open models have been SOTA and that’s for a simple reason :
You don’t need frontier models for all tasks types!!

Qwen models have for months internally totally solved some flows why would I put 5.5-moon-terra-pro for that

[Gergely Orosz @GergelyOrosz]

@vipulgupta2048 open models not considered SOTA

save for GLM-5.2 that many are experimenting with

[Erdal @ErdalToprak]

Yes I’m talking about programming all programming is not do all this one shot slop

[Erdal @ErdalToprak]

Who's using git crypt ?

[Erdal @ErdalToprak]

Hidden on laptop
Fixed left side on external monitor

[Luke @LukeADesign]

Need to know for science, do people have their dock hidden or fixed

[Erdal @ErdalToprak]

temporal vs dbos vs pg_durable vs others ??

[Erdal @ErdalToprak]

glm 5.2 impresses me in ways no other model has ever

it’s a wonderful glimpse at what local intelligence would be like for everyone in a few years

[Erdal @ErdalToprak]

including arbitrarily loosing all conversations in opencode web :(

[dax @thdxr]

a lot of ai coding tools ours included have not been clearing the bar for stability and performance you should demand of a daily driver

james is focused on fixing that and there's some novel things we can do to clear that bar more than we ever have

[Erdal @ErdalToprak]

@paulg Worst football i've seen in a while they truly don't deserve it

[Erdal @ErdalToprak]

pinging @jaywcjlove and @mweinbach that might be interested

[Erdal @ErdalToprak]

Launching ContainerUtility : the missing native macOS app for Apple’s container CLI

It works with machines, containers, images, networks, volumes, registries. It also has diagnostics, activity, runtime health and a menu bar dashboard!

[Erdal @ErdalToprak]

https://t.co/ZQY6hETtXe

[Erdal @ErdalToprak]

@theo yeah overusing geometry reader and lacks coherence on macos apps

[Erdal @ErdalToprak]

I can get codex to go forever with proper long horizon goal that is grounded to verifiable metrics / tasks

[Erdal @ErdalToprak]

That just means you don’t fully use all the tools available
- subagents
- /goal
- knowledge grounding
- alerting the model near compaction to recap everything before compaction

[Aaron Ng @localghost]

Compaction drops Codex's IQ by like 15 points every time. Feels like you have to be really careful with sizing up work.

Part of why Opus & Fable 1M just *feel good* comes from how much you can throw at Claude before the thread feels broken.

[Erdal @ErdalToprak]

There's not a single world where 5.3 codex is ahead of 5.5 what the slop is this

[Mercor @mercor_ai]

Fable 5 is back and we’ve got results for the re-released version on APEX-SWE.

While it did not perform as well as its earlier version from June, the model still significantly outperforms Opus 4.8.

Fable 5 (June): 65.5% Pass@1
Fable 5 (July): 54.8% Pass@1
Opus 4.8: 45.3% Pass@1

This re-release scored about 10 points below the original Fable 5, however it still beat Opus 4.8 by more than 9 points.

[Erdal @ErdalToprak]

The new AI coding paradigm is a tight ecosystem around

- better git allowing remote and local divergences and reconciliation
- better subagent delegation
- saving traces / resolution steps / human in the loop interventions
- distilling proper knowledge from the loop

[Erdal @ErdalToprak]

@solomonneas very cool how do you orchestrate it

[Erdal @ErdalToprak]

Theo and i guess everyone else will discover proxmox lxc agents in 6 months

It's the perfect agent sandbox orchestrator and you can centralize repo state with volume mounts and zfs pools

[Theo - t3.gg @theo]

I have officially moved the majority of my dev work over to Linux. I couldn't be happier. Here's a video about how and why I made the move.