larenna @atomgene

Keep our heads high and burn throught the long night! Joined September 2018

Tweets

161
Followers

13
Following

106
Likes

131

larenna @atomgene

3 days ago

@thsottiaux two hrs past, where's the reset???

0 0 2 545 0

View Details

larenna @atomgene

4 days ago

@shangdu2005 看不懂啊

0 0 0 13 0

View Details

larenna @atomgene

4 days ago

@thsottiaux what can I say, love chatgpt! love openai! love tibo!

0 0 0 4 0

View Details

Codex users: *THIS* is the best way to take advantage of the GPT-5.6 Luna Max capability and 80% price reduction. Treat Luna Max as a 'subagent' by instructing the Sol orchestrator to spin up and monitor a separate thread. Added to 'sol-advisor'. Try it out below 👇

Eric Hauser @ewhauser

2 weeks ago

@daniel_mac8 @pvncher This is so easy to do. Plot out your PRs and then do: Spin off Luna max subthreads to execute. Have them report back then they are complete, review their code, make any corrections, then submit a PR and start a sub thread for the next stack. Finish when all are complete

4 0 32 64K 47

20 20 285 56K 430

View Details

李韭二 @li9292

2 weeks ago

卧槽，刚刚发现了70000个手绘SVG开源库

46 326 3K 345K 5K

View Details

智见AI-大鹏 @zjp1997720

2 weeks ago

🤣卧槽，兄弟们，我这个 Skill 的含金量又被拉爆了！🫡 GPT-5.6 Luna 降价 80%。把推理拉到 XHigh / Max 后，是一个非常适合批量干活的强 Worker。现在大家都在研究一套组合： Sol 做主控 Agent，负责规划、拆解和验收；Luna 做 Worker，负责并行执行。思路没问题，但 Codex 里有个很容易踩的坑：原生 MultiAgent V2 虽然支持指定 Model 和 Effort，但当前原生 spawn_agent 只开放 Sol、Terra，根本叫不出 Luna。我开源的 codex-model-routing-team，正好把这条链路补上了：默认不走原生 spawn，而是通过 Create Thread，为每个 Worker 明确指定 Luna XHigh；高难、高风险任务再升到 Luna Max。Sol 留在主控位统一把关，Luna 并行吃掉执行量。它也不只是一个简单的「模型路由器」。并发上限、数据边界、风险闸门、失败回退、单写者和主 Agent 验收，都已经被固化成一套路由协议。最终得到的是一支真正可控的模型团队：强模型守住判断，便宜又强的模型吞掉执行；质量、成本和速度同时受控。如果你也在折腾「Sol 主控 + Luna Worker」，这个skill给我狠狠的用起来。项目地址： github.com/zjp1997720/zhi… 欢迎star~☀️

Vaibhav (VB) Srivastav @reach_vb

2 weeks ago

You should be Luna-maxxing: > Luna (Max) matches GPT‑5.5 xHigh for roughly 12× less > Terra Max scores 2.6 points higher while costing 45% less

62 53 1K 203K 373

42 43 237 41K 358

View Details

Dan McAteer @daniel_mac8

2 weeks ago

Codex users: here's an amazing way to take advantage of the efficiency and capability of GPT-5.6 Luna Max. It's called 'sol-advisor'. 1. GPT-5.6 Sol High as orchestrator 2. GPT-5.6 Luna Max as implementer for routine tasks 3. GPT-5.6 Terra Max as implementer for complex tasks 4. Fresh GPT-5.6 Sol instance as reviewer A free and open-source plugin for Codex. Installation instructions below 👇 I dare you to install it, and tell me if you hit your weekly usage limits using 'sol-advisor'. I doubt you can.

Dan McAteer @daniel_mac8

2 weeks ago

Codex users: do this right *now*. Max reasoning effort is off by default. Luna at Max reasoning is ~ Sol Medium / Opus 5 Medium level at 1/6th the cost. Change your life today.

186 215 4K 1.1M 3K

120 292 4K 677K 6K

View Details

Adrian Punk @AdrianPunk115

3 weeks ago

x.com/i/article/2081…

39 123 695 261K 1K

View Details

huangserva @servasyy_ai

2 weeks ago

你肯定有过这种时刻：用 Claude Code、Codex、Cursor 开过的 session 数不清，真想找回某次讨论时根本翻不到。只能开新窗口，把项目背景重讲一遍。上一个 session 里磨出来的技术决策、踩过的坑、约定好的写法——一条都没积累下来。我的Mac上光这种记忆就躺着 2832 条：Codex 1704 条、Claude Code 713 条、Cursor 152 条……全是过去几个月一句一句喂出去的，散在 5 个互不相通的工具里。我用 @Memmy_ai 把它们全收回来了。让我惊讶的是，这么好的项目居然开源： GitHub: github.com/MemTensor/memm… 官网：memmy.bot AI 换了，仍然认识同一个你，知道你习惯怎么做事，还能接着上一项任务继续干。一句话说：Let every AI remember the same you——让每个 AI 都记得同一个你。先看这条 3 分钟视频，实测细节在楼下 🧵

116 88 391 80K 712

View Details

Google Research @GoogleResearch

2 weeks ago

Introducing the Science One Framework, an experimental autonomous research prototype that builds verifiable evidence chains. We demonstrate how natively maintaining these chains eliminates hallucinated citations and enables reproducible AI science. Read the blog to learn more: goo.gle/4wzlGWU

50 289 2K 126K 2K

View Details

Amto @XAMTO_AI

2 weeks ago

想象一下，你正在为一道高数题抓耳挠腮，旁边有个懂你的导师，不光能一步步拆解思路，还能顺手出几道变式题巩固，甚至把你之前做错的题都记得清清楚楚，这不再是科幻场景。港大团队开源的DeepTutor，就是一个把这种体验变成现实的AI学习伴侣。一个智能体就能包揽辅导、解题、出题、做研究、画可视化图表，甚至帮你规划完整的学习路径。而且它有个"长期记忆"，能记住你的弱点和进步，再配合好几种不同的知识库检索方式，给出的建议特别贴切。它的技能生态，你可以安装社区里别人写好的教学技能，也能自己定制。源码：github.com/HKUDS/DeepTutor