benjamin zhang

Peace&Love
With no prior deep learning background, using AI in the LLM era to learn deep learning and document my journey.
无深度学习背景，在大模型时代借助 AI 学习深度学习，记录我的学习过程。

[Indie Game Joe @IndieGameJoe]

This indie studio is making a game where you control time by walking left or right.

- Walk right, time moves forwards
- Walk left, time rewinds
- Face front or back and everything freezes

It's called RECUR. This mechanic is genius.

[Deno @deno_land]

Deno 2.9 is released!

› deno desktop
› 2x faster startup, 2x less memory
› min-release-age by default
› css imports
› import npm/pnpm/yarn/bun lockfiles
› deno link/unlink
› deno bundle emits rolled-up d.ts

🦕 full release notes: https://t.co/OlGPZQj7vy

[Pauline P. Narvas @paw_lean]

Reflection of the week: Don’t be around cynical, negative people who only complain for too long. It will make you hate what you love.

[Rockstar Games @RockstarGames]

Pre-orders for Grand Theft Auto VI will officially begin on June 25 on digital storefronts and at other select retailers.

Check out the official cover art, also available as downloadable artwork at https://t.co/XPwC8URCQ4

[Tibo @thsottiaux]

Reminder that you can use the Codex App, CLI and SDK with any open source model, not just with OpenAI models.

https://t.co/spPifB4ck3

[OpenAI Developers @OpenAIDevs]

Use the OpenAI Developers plugin in Codex to build faster with OpenAI tools by setting up API keys, finding the right docs, and debugging along the way.

[Tibo @thsottiaux]

Codex 🤟Ona

Beyond excited to work with Johannes and team to build the future.

[Peter Steinberger 🦞 @steipete]

Here’s your monthly reminder that you shouldn’t be prompting coding agents anymore.

You should be designing loops that prompt your agents.

[benjamin zhang @benjamin_pagar]

@0xZiye 20 太贵了，你看看zero1 或者 eight

[benjamin zhang @benjamin_pagar]

@Russell3402 本就是一家人

[benjamin zhang @benjamin_pagar]

@tindingtin @levelsio 卧槽真的不是AI?

[benjamin zhang @benjamin_pagar]

@gengdaJ 用iCloud是不是也可以，还很安全。

[benjamin zhang @benjamin_pagar]

格局🤟Make OpenAI Great Again!

[Zed @zeddotdev]

You can now use your ChatGPT subscription in the Zed agent, with the same usage and rate limits you benefit from in Codex directly. We're grateful that @openaidevs continues to support subscription-based access for third-party tools, even as others move toward usage-based billing.

[Zed @zeddotdev]

You can now use your ChatGPT subscription in the Zed agent, with the same usage and rate limits you benefit from in Codex directly. We're grateful that @openaidevs continues to support subscription-based access for third-party tools, even as others move toward usage-based billing.

[OpenAI Developers @OpenAIDevs]

Step away from your laptop. Keep building with Codex on your phone.

Codex keeps working on your computer, with your files and project context still in place.

Pocket-sized access. Full Codex working state.

https://t.co/vGsw5wnHnG

[Zain Shah @zan2434]

@eddiejiao_obj @drewocarr @LTXStudio @modal_labs All of this is live! it's early and slow. many of the demos above are sped up/edited, but we can't wait to see what you think. Try it yourself at https://t.co/bcephqPu1c (5/5)

[benjamin zhang @benjamin_pagar]

Context not control

[AYi @AYi_AInotes]

我看完Meta的内部监控备忘录才明白，为啥它投了这么多钱培养了这么多 AI 人才，最后还是在AI赛道掉队了。以及为啥 OpenAI 为什么追不上Anthropic 也可见一斑🙃🙃🙃

Meta最新的内部备忘录，在整个职场和AI圈炸了🤯

4月21日，Meta正式向全美员工推送内部计划，上线MCI追踪工具。
全程记录键鼠输入，鼠标轨迹，随机截取屏幕画面。
对外口径很克制，所有数据仅用于训练办公AI代理，优化自动化操作，不纳入绩效考核。

但这件事的边界早就悄悄模糊了😔

知名投资人 Jason把这件事定义为2026年的职场终极趋势，
在他的逻辑里，这套员工监控体系，不该只用来训AI，
它会成为企业最锋利的管理工具。

精准识别摸鱼的人，剔除混功劳的人，淘汰可被替代的轻中层，

筛选真正创造价值的核心员工，精简冗余架构，倒逼全员产出。

直白的说， AI正在让所有形式的假性工作无处藏身。

我刷了一圈评论区，很多人只看到表层的监控焦虑，忽略了更深层的代价。

其实谷歌多年的团队研究早已证实，高绩效团队的核心底色是心理安全。

也就是说当每一次点击，每一段输入都被实时记录归档，
员工会立刻进入防御状态，放弃试错，放弃创新，放弃跨界协作，
只做可量化，可留痕，不会出错的机械工作。

监控能提升短期效率，却会慢慢腐蚀一家公司的创造力与长期韧性。

因为极致精益的背后往往是人才流失，氛围内卷，和创新枯竭。

而当下AI行业最扎心的差距，也恰恰体现在这里，我们看看 OpenAI 和Anthropic。

2026年，Anthropic在企业营收，客户留存，商业化落地全面反超。

外界总以为，差距卡在模型能力、算力储备或是融资规模。

但事实恰恰相反，
真正拉开鸿沟的，是两家公司完全不同的组织底色。

OpenAI一路狂奔，频繁摇摆，治理结构反复动荡，
高速追热度，追流量，追消费者市场规模。
长期陷入内耗与人才流失，安全团队与商业化路线持续撕裂。

@AnthropicAI 从诞生之初，就带着稳定的长期主义基因。
宪政AI刻入企业文化，长期信托架构兜底发展。
温和审慎的节奏，稳定的决策逻辑，极高的员工心理安全感。
也正因如此，它持续吸纳顶尖研究人才，牢牢拿下高毛利企业市场。

说白了， AI下半场的竞争，早就不是单纯的模型比拼，
已经分化成了两条完全相反的道路。

一边是Meta式的效率逻辑，@Meta
用数据监控压缩人力成本，用自动化替代重复岗位，追求短期利润最大化。

另一边，是Anthropic的长期逻辑。
用稳定治理守住信任，用人文底色留住顶尖人才，靠可靠与合规拿下企业市场。

AI会淘汰低效的工作，这是必然趋势。
但靠监视与压榨换来的效率，我觉得注定走不远。

未来能站稳脚跟的企业，最终拼的不该是极致管控，
应该是在自动化浪潮里，守住人本身的创造力与价值。

[benjamin zhang @benjamin_pagar]

?? @bcherny

[benjamin zhang @benjamin_pagar]

Claude is down ....

[benjamin zhang @benjamin_pagar]

Opus 4.6 在软件工程的架构和技术方案的设计确实要远超 GPT 5.4。怒冲100刀 + GPT 切换到100刀模式，舒服了。