Erik Gao

[Kun Chen @kunchenguid]

pro tip - when you use OpenAI's gpt models in Codex, it uses a server-side encrypted compaction that seems to work better than anything else out there, which allows Codex to just keep hammering on long running tasks like there's infinite context window

that's great, but if you run gpt in other harnesses like Pi, most of them don't inherit that by default, resulting in worse performance in long running tasks

but - because of how extensible Pi is, i just found this cool extension from @alexisgallagher that enables the same server-side compaction in Pi - https://t.co/WboPSMDHR4

benchmark seems to support the argument that OpenAI server side compaction is indeed superior - https://t.co/mHOV69muL7

so if you are using gpt models in Pi, install that extension to improve long running task performance. firstmate benefits a lot from this

if you are using other harnesses, be aware of this difference and see if you can find a similar solution

[Erik Gao @gaaoyi]

OK I paid $200 for this Fable access and then THERE IS A FABEL GATING BUG IN MAX PLAN ARE YOU SERIOUS RN?@claudeai

[Claude @claudeai]

Beginning July 20, Claude Fable 5 will be included in all Max and Team Premium plans, at 50% of limits.

Pro and Team Standard users will continue to have access to Fable via usage credits, and will receive a one-time $100 credit.

Demand for Fable has been challenging to predict, which is why we rolled it out to subscription plans in stages, extending access several times as we secured additional capacity.

[Erik Gao @gaaoyi]

@ThomasLawDeep @higgsfield_ai Same question, what is K3 and GPT 5.6 doing here?

[Erik Gao @gaaoyi]

I did not expect this kind of honesty @Kimi_Moonshot K3 is a great model from great team

[Erik Gao @gaaoyi]

@peduarte @glazeapp thanks for the transparency

[Erik Gao @gaaoyi]

respect both but Andrew seems mad..

[dax @thdxr]

this part made me lol

i was ready for some big reveal on how jarred was trying to corrupt the honorable Zig Foundation in pursuit of self gain

and it was that he asked for an LSP

[Lilian Weng @lilianweng]

new post on harness engineering for AI self-improvement: https://t.co/ZYvGfVs61k

It is hard to forecast how much the future of RSI will rely on harnesses. Likely harness engineering will evolve in the direction of self-improvement and enable auto-research, and, in turn, smarter models keeps harnesses simple.

Even when many harness improvement get eventually internalized into core model, the need to specify goals and context will not disappear.

[Erik Gao @gaaoyi]

@jaywcjlove 等apple container支持compose了试试

[Erik Gao @gaaoyi]

This, from Claude Design🤌

[Erik Gao @gaaoyi]

I'm stunlocked

[Paul Graham @paulg]

How to Earn a Billion Dollars: https://t.co/WeWBUkKym6

[Erik Gao @gaaoyi]

I'm convinced every big AI model lab have secret dataset/benchmark that are hold private and mean to never seen by AI

[OpenAI @OpenAI]

We’ve been researching new ways for ChatGPT memory to carry context across conversations and keep it useful over time.

Today, that work is rolling out as a more capable memory system in ChatGPT. https://t.co/0MyFKCe2Mu

[Erik Gao @gaaoyi]

@turingou 非常赞同 cloudflare的开发者最近经常分享他们调试code mode的事情 agent sdk支持code mode

[Erik Gao @gaaoyi]

Question: How do I know I should spawn a new codex session for the same task? or I just keep talking to it forever?@OpenAIDevs

[Erik Gao @gaaoyi]

@_MaxBlade what is cloudflare agent token?

[Erik Gao @gaaoyi]

what is perplexity been doing？ haven't been on my timeline for a while

[Erik Gao @gaaoyi]

现在最推荐zed。最近新出了thread terminal就是加强版的conductor。本地+remote同时开，和vscode差不多的文本编辑器+Git体验，提醒（包括remote+tmux状态下也会提醒），diff不太顺手但是够用本来ai写的太多代码也不看..

[Viking @vikingmute]

最近发现这个 Herdr 不错啊，https://t.co/NYxNKipTSe

之前用的都是类似 Superset 和 Conductor，但是 GUI 太重了，有时候想要轻量级的，Herdr 直接在原有终端当中运行，继承配置，提供 tmux 级别的持久化，Agent 状态感知，支持多 Agent 并行，非常轻量级， Rust 二进制。

今天小试了一把，很喜欢，现在各种应用占用资源太多了，有时候需要这种轻量级的解决方案。

还有 Responsive TUI，在手机上看也很友好。（这个概念我是第一次听说，其实很受用）

[Erik Gao @gaaoyi]

Bro any chance the compaction logic gets open sourced? The compaction works so good in codex😭

[Greg Brockman @gdb]

under appreciated that codex is open source

[Erik Gao @gaaoyi]

This make Zed better at what conductor try to do, because Zed have pretty mature ssh remote features, which conductor and superset can’t yet. Now local and remote AI session all integrated with one panel with nice little things like notifications and text editor

[Zed @zeddotdev]

Terminal Threads are live in Zed v1.3.5!

You can now run claude, amp, pi, or any terminal-based workflow as a managed thread in the Threads Sidebar, right next to your other agent threads.

[Erik Gao @gaaoyi]

@waylybaye 持续在用这种方式，说明一只能从AI的代码中发现确实写得不行？