deepsky

[deepsky @OmegAlphApollo]

@mediastorm_en @grok pick me pls

[Tim Pan @mediastorm_en]

Winner announcement in 24 hours! We'll randomly draw 100 eligible winners with help from @grok. Make sure you follow us, like & repost to be in the running. Good luck! 🍀

[deepsky @OmegAlphApollo]

@Mashiro_0619 我比较喜欢 omp，搭配自己的 skill
https://t.co/VPFbhsnsU9

[deepsky @OmegAlphApollo]

@0xkenzolee @grok DONE

[deepsky @OmegAlphApollo]

@NanfordLynn @thsottiaux 🤣

[deepsky @OmegAlphApollo]

@shenpengwdh 用 grok bot email 订阅 newsletter，搜集时事咨询，整理汇总后定时推送给我，定制个人专属信息流

[deepsky @OmegAlphApollo]

@Rakel076 开新会话，避免上下文污染。
也可以在一个会话中工作，跟主 agent 沟通，让它自行分派子 agent 工作，可以参考我的两个 skill

https://t.co/VPFbhsnsU9

[deepsky @OmegAlphApollo]

Tibo’s 28-day challenge visual tracker.
Improvements vs resets so far — take a look 👇

[deepsky @OmegAlphApollo]

Tibo 的 28 天更新挑战的可视化追踪。
当前的改进vs重置情况，瞅一眼👇

[deepsky @OmegAlphApollo]

Love it. 🥳

All my newsletters now go to my Grok email.
Grok reads, filters, and summarizes them for me on my terms.

Goodbye information overload.
Hello only the stuff that actually moves the needle.

[Grok Bot @bot]

Grok Bot now has its own email.

Bot can use it to sign up for services, contact businesses for you, or schedule time with someone.

[deepsky @OmegAlphApollo]

你是不是也这样？😩
让AI写代码 → 它直��开写，没有计划
跑到一半上下文爆了，关键信息全丢 🤯
多个代理一起改，文件互相覆盖 💥
最后它说“做完了”，你却不知道到底验过没有 ✅❌

Orchestrate v1.0.0 就是来解决这个的。

主 Agent 只负责计划和最终验收
子 Agent 去调查、实现、集成、修复

✅ 所有权清晰
✅ 并行可控（不打架）
✅ 有证据才算完成
✅ 上下文压缩也不丢状态
直接调 orchestrate 就能用 🚀
搭配我的 three-loop-workflow 效果更好
（Plan → Build → Close + 独立审查）

👉 https://t.co/FeoOPansl1
👉 https://t.co/VbdyWAYhzE

[deepsky @OmegAlphApollo]

Sound familiar? 😩
You ask an AI to code → it jumps straight in with no plan
Halfway through, context explodes and critical info is gone 🤯
Multiple agents edit the same files and overwrite each other 💥
It says “done” — but you have no real evidence it was actually verified ✅❌

Orchestrate v1.0.0 fixes this.

Parent agent owns the plan + final acceptance
Subagents investigate, implement, integrate, and fix

✅ Clear ownership
✅ Controlled parallel work (no conflicts)
✅ Evidence of completion required
✅ Context survives compaction
Just say orchestrate to use it 🚀
Pairs especially well with my three-loop-workflow skill
(Plan → Build → Close + independent review)

👉 https://t.co/FeoOPansl1
👉 https://t.co/VbdyWAYhzE

[deepsky @OmegAlphApollo]

lol

[deepsky @OmegAlphApollo]

@RiverAi7z @leto_bao 不能更赞同，我现在就是这么玩，再搭配一台云服务器，非常爽

[deepsky @OmegAlphApollo]

totally agree!

I have set STE as a must follow rule in my CLAUDE.md/AGENTS.md several months ago, and enjoy it a lot.

If you have time, I recommend you to download from https://t.co/fmSIK1etMv , and read the document by yourself, it’s free and useful.

[Andrej Karpathy @karpathy]

We'll be spending a lot more time trying to understand the outputs of language models. A few thoughts, tips & tricks:

Writing. Something I've had success with: Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent. But even better:

Diagrams / images. Instead of writing, ask your LLM to create a diagram. These can be a lot easier to process, parse, and understand. But even better:

Web pages. Ask for output "in HTML" to get a beautiful, interactive webpage. LLMs are getting really good at frontend and can create beautiful experiences, animations, etc. But even better:

Explainer videos. The output format I am most bullish on is fully custom / bespoke explainer videos generated on any arbitrary topic. Experiment with things like "Create a 3b1b style video explainer on X. Use my ElevenLabs API key for audio narration". (you'd need an API key for the latter or you can ask your LLM to find you decent free alternatives that use your local compute). This is actually starting to work!

In summary:
- As LLMs get better, they will do more and more of the legwork autonomously, and a lot more of our work will rise up the abstractions into oversight and understanding.
- Luckily, LLMs can help here too because as intelligence and code are increasingly abundant, you can ask for large, custom, discardable software artifacts (e.g. web apps, video explainers) that would have never made sense to create before. Push the boundaries here and you'll be surprised.

[deepsky @OmegAlphApollo]

Damn bro

[deepsky @OmegAlphApollo]

Looking for an x86, portable, Linux-friendly laptop.
No ARM. Prefer something light that just works on Ubuntu/Fedora (suspend, Wi-Fi, fingerprint).
What are you actually using in 2026?

[deepsky @OmegAlphApollo]

@trq212 I never use it. In most cases, it just makes me, the human, feel better — it doesn’t make agents perform any better.

[deepsky @OmegAlphApollo]

This isn’t just a Fable trick.

Same split works anywhere the best model is scarce: Claude Code, multi-agent setups, research pipelines, support bots, internal tools.

Put the expensive model on decompose / assign / review. Put cheaper agents on the labor.

The scarce model lasts longer because it stops doing work it didn’t need to touch.

[deepsky @OmegAlphApollo]

Fable is the model you actually want.

It’s also the one that dies first.

Two real sessions and the 5-hour bar is gone. Sometimes 30 minutes. Then you’re on Opus, pretending that’s fine.

[deepsky @OmegAlphApollo]

Paste this at the top:

Work only as an orchestrator. Do not do the actual work. Split the request into practical tasks, assign each to a suitable model, then review and accept the results.

You keep the judgment. You stop burning the quota on drafts.