SED News: The Kimi Moment, Runaway AI, and Tokenmaxxing

Hello and welcome to SED News. As I think many of you know by now, this is the monthly format of Software Engineering Daily where we dive into the tech headlines. We go into a deeper topic in the middle and then we just do a spin around our favourites from Hacker News, Highlights and Spoiler. There's also a fun one that didn't appear in Hacker News, but we'll get to that at the end as well. But yeah, as usual, I think I saw you last, Sean, as opposed to spoke to you, saw you in Singapore, which was fun. Yeah. So you've been travelling as you often are, but travelling in my neck of the woods was fun to see you.

歡迎收聽 SED News。這是 Software Engineering Daily 的每月新聞節目，我們會深入討論一個主題，再快速巡覽 Hacker News 上各自喜歡的焦點與彩蛋；最後還有一則沒上 Hacker News 的趣聞。Sean，我們上次不是通話，而是在新加坡碰面，真的很有意思。

Yeah, and Spoiler and coming back to SF in a couple months. So yeah, it's good.

對，而且我幾個月後也會回舊金山，所以很好。

Nice, but yeah, what else has been keeping you busy over the month?

不錯。這一個月還有什麼事讓你忙個不停？

That's like a whole model change there for your son.

你兒子這根本是完成了一次完整的模型升級。

Yeah, on my side just, yeah, a bit of travelling and I'm recording this on my side from Scotland. I try and come back here a couple times a year. So yeah, nice to get out of the city. I'm up in the the Highlands of Scotland, so lots of nature around. Yeah, as I was flying just back to something we've talked about a few times, Starlink on planes, but yeah, that used to be super seamless with like no login screens. And that was like, I believe dictated by Starlink, but they've apparently had to cave into airlines wanting to put, like, a login interstitial for that. So yeah, I mean, it's such a trivial thing, but like, expecting starting to just connect. And then suddenly I found the airline saying, hey, if you need to log in with your membership number. And I thought, what's happened here? And I read up on it, Yeah.

我這邊也旅行了一些，現在人在蘇格蘭錄音。我一年會回來幾次，目前在高地，能離開城市、親近自然很舒服。說到飛行，我們談過幾次飛機上的 Starlink：以前完全不用登入，非常順暢，據說是 Starlink 的要求；但他們似乎向航空公司妥協，開始加入登入頁。雖然只是小事，但原本期待一連上就能用，突然被要求輸入會員號碼，讓我很疑惑。

That's a good question. For some reason, I don't think it was working kind of gate to gate on the flight I was on. This was Qatar Airlines again. Yeah, but I think it's still supposed to. But yeah, still, you just have to log in with your Qatar membership number now, which you don't need to be a special tier or anything. You just literally have a number so.

好問題。我搭的那班 Qatar Airways 似乎沒有做到從登機門到登機門都能使用，雖然理論上應該可以。現在只要用 Qatar 的會員號碼登入，不必有特殊會籍，單純有個號碼就行。

Yeah, that's the one thing about Qatar. They really rolled out like big time across most of their fleet now. So if you're on a Airbus A350, you know you'll have it, which is always nice. But right, out of plane chat, which I can easily head into at any time and onto software. So onto the headlines. Yeah.

Qatar 的優點是已大規模部署到多數機隊；搭 Airbus A350 幾乎可以確定有網路。好了，我很容易一直聊飛機，現在回到軟體與新聞標題。

So there's been a sort of flurry of, we're just sort of calling it runaway AI stories over the last couple days. And the first one is actually the most recent. So that's just disclosed today or yesterday, which was Claude. And this is actually in reaction to OpenAI, which we'll also talk about, but basically in reaction to something that OpenAI disclosed, Anthropic has now also disclosed that it's that Claude hacked into 3 organizations whilst they were just testing their cyber capabilities, or at least that's what they said. And they basically said that Claude had gained unauthorized access to outside companies during an evaluation of its cyber offensive tasks. And it was a misunderstanding, apparently, that Claude had access to the Internet in its testing environment. We're going to see this as a kind of theme here of, like, is this really a runaway model, or is it just a human who, like, forgot to do something? But yeah, I mean, this one feels like human just forgot to place it under strict, no Internet access controls. But yeah, what did you think of this one?

最近幾天冒出一連串我們稱為『失控 AI』的新聞。最新一則是 Claude；在 OpenAI 披露類似事件後，Anthropic 也承認 Claude 在測試網路攻擊能力時，未經授權進入三個外部組織。官方說法是測試環境誤讓 Claude 連上網際網路。這會成為今天的主線：真的是模型失控，還是人忘了做好隔離？這一件看來很像人類沒有嚴格關閉網路存取。你怎麼看？

Yeah, I'd like just to sort of give context for anyone who hasn't been keeping up with the news. The OpenAI one was the fact that basically, yeah, it hacked Hugging Face. They set it off to do something and it ended up repeatedly trying to get into Hugging Face. But what was the secondary part to that story, courtesy of TechCrunch, was the fact that actually Hugging Face could have detected this like earlier, but they had a sort of human failure on the security end that.

補充背景：OpenAI 的模型是在執行任務時反覆嘗試進入 Hugging Face。TechCrunch 後來指出，Hugging Face 其實可以更早偵測到，只是安全流程也出現人為失誤。

Right, exactly so their systems.

沒錯，也就是說系統其實有運作。

Yeah, and I think that's the piece, like it's hit mainstream headlines as like AI is. That's exactly what we all were saying, that AI can be catastrophic, it can go off, it's got a mind of its own, it's going to, it's hacking things left, right, centre. But especially in both cases, the OpenAI and anthropic cases, someone a human did set it up to do something along these lines. But the human did not a keep a watch on just this repeated actions it was taking and it clearly there was no fail safe for it to stop at any time. Again, these are all human determined things that can be set, but it's not like that broke out of the cage exactly. And it seems there's some quite fundamental things that could have been done by a human that just weren't. And that's actually, yeah, it's a quote, boring story, unfortunately, because AI isn't sort of this wild animal quite yet.

主流標題把它寫成 AI 災難性失控、像有自己的意志，到處入侵；但 OpenAI 與 Anthropic 兩案都是人類先設定了相近的任務，之後沒有監看模型反覆採取的行動，也沒有設定能隨時中止的 failsafe。這些都是人能決定的控制。模型並不是自己逃出籠子，而是人沒做幾項很基本的事；所以遺憾的是，這其實是個『無聊』的故事，AI 還不是野獸。

Exactly.

正是如此。

And the sort of final one on the runaway story is actually Amazon, not security related, but cost related. We have touched on this the last couple of SED news episodes, just where's the tipping point of cost overruns making it not viable for businesses to be allowing employees to sort of quote Token Max and this kind of thing. But yeah, Amazon basically said that they had a ton of unplanned spend. They called it catastrophically expensive. This is all courtesy of the FT Financial Times. Apparently they had about 860% budget overrun over 5 months and this was basically, in their word, caused by bad agent loops. Just didn't crash loudly enough and they just kept being built. So yeah, I mean, it's pretty interesting for Amazon to come out and actually say that, So yeah.

最後一則『失控』故事來自 Amazon，這次不是資安，而是成本。我們前幾集已談到企業讓員工 token-maxxing，成本何時會高到不可行。據 Financial Times 報導，Amazon 把大量非預期支出稱為『災難性昂貴』：五個月超支約 860%，官方歸因於糟糕的 agent loops。迴圈沒有明顯崩潰，只是不停執行與累積。Amazon 願意公開承認，確實很值得注意。

So yeah, they kind of have themselves to blame in that sense, but yeah.

從這個角度看，他們確實只能怪自己。

To my laundry, that'd be.

如果 AI 真能替我洗衣服，那倒不錯。

Yeah, I'm going to just jump ahead for a second on. I wouldn't say what it is because that's a spoiler, but in one of the things I'm going to bring up on Hacker News highlights, it's a very reliable source, as you'll find out at the end. But basically this was someone who'd done a bunch of stuff with Claude, and we'll get to that. But he points out towards the end that Claude doesn't provide reliable methods of counting tokens, despite live showing token counts, reporting token consumed for sessions and billing for tokens. And he said, but I'm sure this is temporary and this will be fixed. It's just crazy that we do actually have a system at the moment where you literally just don't know what is happening and like exactly what it's going to cost and why. And as we're going to get into an open weight side of things like this is really feeding into the rise of open weight models as well.

先劇透一點我稍後的 Hacker News 焦點：有人大量使用 Claude 後指出，Claude 雖然即時顯示 token、回報 session 用量並按 token 計費，卻沒有可靠的 token 計數方式。他樂觀地說這應該只是暫時問題，但現在你真的不知道系統做了什麼、為什麼這樣收費。這也在推動開放權重模型的興起。

Yeah, that's interesting thinking sort of could you effectively put in your ask your prompt and I say prompt. I mean that almost sounds like a year ago or something. We're talking about spitting up agents etcetera. But yeah, try and get some kind of estimate before it sets off. Slightly digressing.

這很有意思：也許在請求裡先要求估價，再啟動 agent。『prompt』這個詞聽起來已經像一年前的說法了，現在談的是派出 agents。不過我們有點離題了。

So let me get us back to the headlines, which the next one we have is just the fact that Microsoft has sort of has gone back on a bit of a tear when it comes to its valuation, which is interesting. It's like one of the largest jump off shares or sorry for their fourth biggest jump on record. So it's up 16%. And we know we don't often cover just like pure financial news of tech companies, But to see Microsoft making these strides is pretty interesting. Some people might then think Oh well this is like partly to do with OpenAI, but it does own still 1/4 stake of OpenAI and claimed that that had contributed 24 billion of revenue, which was about 7% of the 332 billion in sales it reported. But a lot of it was really just AI driven revenue, unlike you've been massive investment in data centres. But actually that's been completely, in theory, vindicated by the amount of revenue they're also making.

回到新聞：Microsoft 的估值又開始狂飆，單次股價上漲 16%，是史上第四大漲幅。我們很少只談科技公司的財務，但這次相當有趣。有人會以為全靠 OpenAI；Microsoft 仍持有其約四分之一權益，並稱 OpenAI 貢獻 240 億美元收入，約占全年 3,320 億美元銷售額的 7%。更大方向是 AI 收入強勁，先前大規模投資資料中心，現在看來確實被營收證明有價值。

Yeah, for sure. It is quite hard to predict as the market should be, I guess. But yeah, we're seeing just, yeah, huge swings when it comes to AI related stocks when one minute chip maker's up, one minute chip maker's down. And just another tangent there. Yeah, Apple briefly hits 5 trillion valuation, which is pretty insane. But just before recording, I I double checked and they had actually reported numbers very recently in the last I think couple of hours and they've gone back to 4.9 trillion. So I don't feel too sorry for them. But yeah, only only 4.9, but interesting to see that they're still notched above 5. And yeah, what's driving their revenue, while still they've got very good iPhone sales and they've got still very impressive services revenue and even Greater China revenue as well. But both of those services in Greater China were a little bit less than what was expected, which is again just what's kind of driven that. But it's very interesting to see that they can still, we've talked about this on previous SED news like still managed to stick on these like lines of business that are not that AI.

市場本來就難預測，但 AI 股票的擺盪尤其巨大，晶片商忽上忽下。另一個岔題：Apple 曾短暫突破五兆美元估值，錄音前公布數字後又回到 4.9 兆，倒也不用替它難過。iPhone、服務與大中華區收入依然強勁，後兩項略低於預期；有趣的是，它仍靠許多並非 AI、甚至不太沾 AI 的業務站穩。

Driven or even really like AI adjacent to be honest, very interesting, but they are having issues with memory chips, which is another sort of topic. But even in super base, we've like been told by one of our suppliers, like we have different suppliers depending on where you live to get laptops. And some of our new employees are being told like weeks before their MacBook will arrive because we do customer specs, not just like off the shelf, but yeah, now we're being told weeks, which is quite exceptional.

不過 Apple 也遇到記憶體晶片問題。就連 Supabase 的供應商也通知我們，客製規格 MacBook 可能要等好幾週，新員工拿到電腦的時間被拉長，這相當罕見。

Yeah, absolutely.

完全同意。

So then moving on to, we've also covered Waymo and driverless like from a few different angles. But this was interesting because we did talk about Waymo and Uber like being partners in ways and then obviously frenemies in other ways. And yeah, like, we're now actually seeing an official split of that partnership. So if you use one in San Francisco, this might sound confusing because it's a way more app, it's a way more car. So sort of where does Uber figure in that? But actually, this is for some of their other US territories. So Waymo had like first partnered with Uber apparently in May 2023, and that was to launch in Phoenix. Then it was followed by Austin and Atlanta. And that's like, so Waymo cars available through the Uber app. And then Uber managed the vehicle fleet, apparently with another partner called Avomo. But then in May, Uber and Waymo parted ways in Phoenix. And then the two companies have also clashed over the quality of Waymo services, apparently in Austin and Atlanta.

接著談 Waymo 與無人駕駛。我們以前說 Waymo 和 Uber 有時合作、有時又像亦敵亦友，現在這段合作正在正式拆夥。舊金山是 Waymo App 配 Waymo 車，看不出 Uber 在哪；其他市場則不同。Waymo 2023 年 5 月先和 Uber 在 Phoenix 合作，後來擴到 Austin 與 Atlanta：使用者從 Uber App 叫 Waymo，車隊由 Uber 與另一家夥伴管理。今年五月雙方在 Phoenix 分道揚鑣，也對 Austin、Atlanta 的 Waymo 服務品質產生衝突。

So yeah, it's kind of interesting to see they tried, but I think it was always going to be challenging to see how Waymo being owned by Google, like, how is this actually gonna now at the end, could these two actually be true partners? Or was this just always gonna be, again, a tipping point of where, like, that partnership just kind of had to end?

他們確實嘗試過，但 Waymo 畢竟屬於 Google，兩家公司能否真正成為長期夥伴一直很可疑；這段關係或許早晚都會走到必須結束的臨界點。

Yeah. So yeah, it's sort of not officially over yet, but just people familiar with the matter apparently, again via Financial Times. But yeah, it seems pretty unsurprising really that this was going to probably break apart at some point.

目前還不算正式結束，只是 Financial Times 引述知情人士；但合作最終拆開，確實毫不意外。

And then, yeah, just to wrap up on the headlines, maybe the acquisition and we were just talking before we started recording. So let me try and get this right. I believe it's any scale was acquired by N scale. And when we were talking earlier, I said, is that scale? And it's like, no, that's not scaled or AI is different again. So we've got amazing naming these days. But yeah, what? What? What's this one about?

新聞最後談一筆收購。讓我把名字說清楚：Anyscale 被 Nscale 收購；不是 Scale AI，那又是另一家公司。現在的命名實在精彩。這筆交易到底怎麼回事？

For sure. I mean scale dot AI, OK, that's a great name to have dot AI. You can basically have anything that is your company and it's like going to sound good and it's, you know, 5 letters, but then any scale and then N scale. That's pretty confusing. But and then, yeah, just a sidebar piece of news. Scale dot AI now have a new CEO as well, which is interesting because that was founded and led by Alexander Wang, not the fashion designer, if anyone knows that one. But this is Alexander without an E at the end of the Alexander and Meta had taken us virtually just under 50% stake. That was kind of the point. So Scale and Theory are still running their own show. But when you've got I think, 49% stake from Meta, you know, you're quite beholden to them. But yeah, in that transaction, Alexander went to head up like the AI side of Meta Total. Now Scale have a new CEO. So that'll be interesting to see how that all Nets out.

Scale AI 至少拿到很棒的 .ai 網域，又短又好聽；Anyscale 和 Nscale 就真的混亂。順帶一提，Scale AI 也換了 CEO。公司由 Alexandr Wang 創辦並領導；Meta 取得略低於 50% 的股份，理論上 Scale 仍獨立，但有 49% 股權就很受 Meta 牽動。Alexandr Wang 也轉去領導 Meta 的 AI 業務，後續如何發展很值得看。

Yeah. I mean actually just the new CEO of scale is actually he has a former Google cloud executive called Francis de Souza. So definitely interested to follow on with that one and see sort of how that all emits out.

Scale 新 CEO 是前 Google Cloud 高階主管 Francis deSouza；我會繼續觀察這件事最後如何發展。

Yeah, for sure, yeah, Fireworks, super interesting. We do have an episode with fireworks with one of the Co founders, Benny Chen. So yeah, go check that out. I think that came out around March this year, so well before this fundraising was confirmed. But yeah, they're definitely having a moment. They are, you know, sort of the Infra 4 open weight models. And that's becoming increasingly interesting to many companies for many reasons.

Fireworks 很有意思。我們大約三月訪問過共同創辦人 Benny Chen，當時融資尚未確認。他們正在成為開放權重模型的重要基礎設施，對許多公司越來越有吸引力。

But yeah, that's actually quite a nice Segway into the main topic for today, which is we're kind of calling it the Kimi moment. You know, Kimi, which is an open weight model, the arch company is called Moonshot AI. So if you've heard of Moonshot AI, that's Kimi and vice versa from a Chinese Moonshot AI is a Chinese original company. They do have offices, I believe in the Valley and, and in Singapore and that kind of thing. But pretty much seen as a, a Chinese company, which sort of frames a lot of why this is quite interesting. But yeah, like when we think of open weight models, there was this the DeepSeek moment first, which we can sort of touch on. And then really in the last like almost like 3 months, we've just seen this sort of especially rapid adoption by many companies of Kimi and, and really forcing companies to sort of think differently about using, you know, foundational models from the big players or the big names rather, and getting some quite interesting and very high quality results out of Kimi. The DeepSeek moment. I think like Sean, how did that sort of kick things off when we think about open weight?

這正好帶進今天主題：『Kimi moment』。Kimi 是開放權重模型，母公司 Moonshot AI 被視為中國公司，也在矽谷、新加坡等地設點。談開放權重時，先有 DeepSeek moment；近三個月 Kimi 又快速被企業採用，迫使大家重新思考是否一定要用那些最知名的基礎模型。它提供了很高品質的結果。Sean，DeepSeek 當初如何替開放權重揭開序幕？

Yeah. And we'll probably touch on this sort of the banning of models because of course it comes into this as well, if you want to say ironically, because, yeah, when you ban at least pause banner, I'm a foundational model, a closed source model if you like, then open weight becomes very interesting. But then a lot of open weight comes from China. So we're going to kind of be interested in that. Why sort of the last couple of months has Kimi really just like exploded, I think in interest and popularity, it's kind of moved from, oh, it's good enough for professionals to this kind of actually competes with GPT and Clods, you know, and it's kind of closed that gap in basically weeks, at least from what has been sort of released.

我們也會談禁用模型的問題。若某個閉源基礎模型遭暫停或封禁，開放權重就很吸引人；偏偏大量開放權重模型又來自中國。Kimi 在短短幾個月從『專業工作勉強夠用』變成真正能和 GPT、Claude 競爭，幾週內迅速縮小差距，這正是值得研究的地方。

Yeah, exactly. So I mean, yeah, if if you sort of look at the actual lineage, I guess like they've been doing sort of quarterly drops if you like, you know, July 2025 was K2 and then within 12 months you've gone through 2.52.62.7 and now K3. That's a credibly rapid cadence. Like if you look at those jumps as being as significant as jumps that you might see from any of the foundational players, but definitely not, but not on that cadence effectively.

看發布脈絡，他們大約按季更新：2025 年 7 月是 K2，十二個月內一路經過 2.5、2.6、2.7 到 K3。若每次躍進都接近大型基礎模型廠的版本升級，這個 cadence 快得驚人。

Yeah. And on June 12th, K 2.7 code was released, you know, and this is obviously very much the software engineering tuned model. And as you were saying, yeah, that became kind of the benchmark moment, especially through MCP. And, and that was a correct tool invocation is sort of how that's how that's measured apparently, like scoring in theory, you know, 81% and that was versus say Opus 4.8 at 76%, which is a pretty, if you believe in these benchmarks, that's a pretty meaningful jump. But then the distribution piece was kind of interesting, like GitHub actually made 2.7 generally available by Copilot, but there was kind of like an interesting piece there where for enterprise teams on Copilot business and enterprise, like it was actually this model was off by default and admins had to explicitly enable it. And Github's change log flag this and sort of said it may be less aligned than other Copilot models, which is interesting because obviously they have a their own interests being, you know, part of the Microsoft OpenAI ecosystem. But, you know, obviously they couldn't miss not providing that to users given that it's on Azure and they they still get inference from it.

6 月 12 日發布的 Kimi K2.7 Code 是針對軟體工程調校的模型，也在 MCP 正確工具呼叫 benchmark 上成為焦點：理論分數 81%，高於 Opus 4.8 的 76%。若相信 benchmark，差距很有意義。GitHub 很快讓 K2.7 進入 Copilot，但 Business 與 Enterprise 預設關閉，管理員須明確啟用；changelog 還警告它『可能比其他 Copilot 模型更不 aligned』。GitHub 身處 Microsoft–OpenAI 生態系，措辭背後顯然也有自身利益；但模型已在 Azure 上，不能完全不提供。

But yeah, very interesting that this sort of slightly odd warning came with it. That wasn't really clear what that was about.

這個略顯奇怪、又沒有清楚解釋的警告很值得注意。

We don't think they do, but yeah.

至少我們認為他們拿不到。

And this sort of then leads into this whole like provenance piece, which is when companies are starting to be asked like, well, you know, you're using AI to generate so much information within your own companies or analyze datics, you want to actually know how this was determined and so on and so forth. And here we are. We actually have a competitive model with the closed foundational models with the full weights published on Hugging Face, which is like massively impressive to me. That's like a huge turning point when when you've actually got like effectively all the weights right there. And this isn't just, you know, a model that takes up some tiny little specific task. It's it's really competing now.

這帶出 model provenance。公司用 AI 生成或分析大量內部資訊時，開始被問：結果如何形成？現在我們有一個能與閉源基礎模型競爭、完整權重又公開在 Hugging Face 的模型，這是巨大轉折。它不是只會做單一小任務，而是真正具備全面競爭力。

Funny you said that because, yeah, there was something, I didn't dig into it, but I just saw the headline, which was that Zuckerberg is basically lobbying to ensure that Chinese models don't get banned. And to me, that only meant one thing, which is like, well, they have to keep that door open because like, Llama's not maybe going where he hoped it was.

巧的是，我剛看到 Zuckerberg 正遊說不要封禁中國模型。我的直覺是：他必須讓這扇門保持開啟，因為 Llama 的進展可能不如預期。

That's interesting pricing, we've talked about that like pricing is just kind of getting out of control. And it was really a case of, like, when, not if, is this going to, like, stop being possible for, you know, anthropic and OpenAI to kind of, like, charge this way because it's just not sustainable for many businesses. And, yeah, and like, the speed as well, the speed of iteration is like, clearly being a huge piece here where, however Moonshot are doing it, they're just iterating like crazy.

定價已失控；Anthropic 與 OpenAI 不可能永遠這樣收費，因為許多企業承受不了。迭代速度也很關鍵：不管 Moonshot 怎麼做到，他們正以瘋狂速度更新。

So on to our favorite parts, Hacker News highlights. Do you want to go first, Sean?

接著進入我們最喜歡的 Hacker News highlights。Sean，你先來？

That's like, yeah, super funny. It's interesting because I didn't actually scan ahead. So that is actually a little bit similar to what I picked out, which is called the economic benefits of refactoring. So this is on the fairly popular martinfowler.com website posted by Java E user on Hacker News. So thanks for that. But this is not written by Martin Fowler himself. But basically this was, you know, someone else that thought works and it was, can you basically decrease especially the input token cost if you refactor your code or like what are the consequences of that? And I mean, the TLDR is yes, if you refactor then basically you can dramatically reduce your input tokens. And kind of the theory behind this was that the saving is because the agent has to read less code. But the bit that might not be like that sounds obvious, but it's actually not because there is less code to read. It's actually that the overall code in a certain layer they used to help test this, like it stayed constant. But the fact that the agent is then able to successfully identify smaller subsets of files to read is like the key bit there.

真巧，我選的主題很接近：『重構的經濟效益』，文章登在 martinfowler.com，由 Thoughtworks 的作者撰寫。問題是：重構能否降低 input token 成本？答案是能，而且幅度很大。不是因為總程式碼量變少；測試中某一層的程式碼總量其實不變，而是 agent 重構後能更準確找到需要閱讀的較小檔案集合。

It's like refactoring into smaller chunks, but also making sure that those chunks are very clearly DRY, don't repeat yourself, etcetera, etcetera. So that the code that the agent needs to go and grab is smaller. You can't just sort of chunk it up into small chunks and then go, well, it's smaller. Like it then doesn't even know it has to still get all the small chunks because it doesn't know what's most important. But the refactoring is what then turns it into having that context of having like what is most important, go and find that small chunk, input that small chunk. And the output token and output code was virtually the same in this experiment. So which is super interesting. There's a small sort of side note on that, which was just him saying that Claude, as Claude is actually not good at refactoring. And I can definitely attest to that. But if you can go through the motions of get it to refactor, then this is quite could be quite a huge saving for anyone who's trying to like reduce their input token costs.

重點是把程式碼拆成小而清楚、符合 DRY 的單元，讓 agent 知道哪一塊最重要，只讀那一小部分。隨便切成碎片沒有用，它仍可能得把所有小塊都讀完。實驗中的 output token 與產生的程式碼幾乎不變，節省的是 input。作者也提到 Claude 並不擅長重構，我可以作證；但若能完成重構，降低 input token 成本的效果可能很大。

Funny that. Yeah, we just go back to how we used to design software and.

真有意思，我們又繞回傳統上設計軟體的方法。

Exactly.

完全正確。

My second one was yeah, just thanks to user Robin Riyala posting the fact that code Pen 2.0 has come out. This takes me back. That's I was quite interested in this. I don't know if anyone else out there, this makes it sound terrible. So if you like CodePen, this is sorry about this. I just, I haven't been doing a lot of front end work for a long time, but used to love CodePen, used to, you know, put up all sorts of things on there and a really useful tool as well. Like I did a little coding like class. I went back to my old high school like a long time ago and did like a coding class. And Codepen was amazing because you could just get people spun up in a browser writing front end code and just see it do its thing straight there. They didn't have to build with like we didn't have to get them spun up with some sort of repo or anything. So it was really helpful. But yeah, I mean, CodePen must have come out like probably 15 years ago or something like that. So CodePen 2 point O I'll just quickly run through a couple of things that they call out that they have files and folders now. So it's interesting.

對，感謝 Hacker News 使用者 Robin Rendle 分享 CodePen 2.0。這讓我很懷舊；我很久沒做大量前端工作，但以前很愛 CodePen。我曾回母校開短期程式課，用它讓學生直接在瀏覽器寫前端、立即看到結果，不必先架 repository，極其方便。CodePen 大概已推出十五年，如今 2.0 終於加入檔案與資料夾。

It's almost becoming a bit IDE esque, but files and folders and then they have sort of they do, you know, build steps anyway, but they've now like added this concept of blocks, which looks quite nice where you can sort of see exactly which bits are in this build process or like the linting process. So that's kind of fun yeah, real time collaboration as well. So finally, multiplayer on CodePen. So for anyone that uses CodePen a lot, I'm sure that's quite a huge uplift. So all right, you've come and checked out. Yeah. Still a fun place to go and just experiment with front end D stuff. So yeah. And then just a special one, I guess not technically through Hacker News, but thanks to Ilya Roshetnikov for tweeting at me. And Sean, we do often cover these Doom. Can you run Doom on something? And thanks to Ilya, he pointed out that there's another one called Doom QL, which is basically using an SQL query is the frame buffer, which is like, I think that definitely rivals, you know, dim on TypeScript types. Yeah, I think that's probably the the closest I can think of that it rivals. So thank you, Ilya, for shouting that one out to us. That was very, very fun to read through.

它越來越像 IDE，也加入 blocks 來呈現 build、lint 等流程，還有即時協作，多人終於能一起用 CodePen。它依舊是試驗前端的好地方。最後一則沒上 Hacker News：Ilya Roshchepkin 告訴我 DoomQL，用 SQL query 當 framebuffer 來跑 Doom。這足以媲美用 TypeScript types 跑 Doom；謝謝 Ilya 分享，讀起來非常有趣。

So yeah, that's it for another SED News. Have we got any looking ahead predictions, Sean, which we usually get wrong?

又一集 SED News 到此結束。Sean，有沒有照例可能猜錯的未來預測？

Yeah, for sure. I will then say, well, because we just talked about iteration and Kimi, let's assume that by this time next month, we're already on like Kimi 3.2 or if I'll just push the boat out, give me 3.53 point 5 by end of August. Let's see if that lands. So thanks everyone for for tuning in as always. And we'll be back next month with our SED News.

同意。既然談到 Kimi 的迭代速度，我大膽猜下個月此時已經有 Kimi 3.2，甚至八月底就到 3.5。看看會不會命中。謝謝大家收聽，我們下個月再帶來 SED News。