tkyshat retweeted
Garry Tan, the CEO of Y Combinator, open-sourced 55 ready-made agents for Claude Code
Each agent is a slash command in plain Markdown with its own role and rules. 134k stars, MIT.
What is inside:
> 7 for browser work and scraping
> 6 for product planning
> 6 for design
> 6 for memory and retros
> 5 for code review and debugging
> 5 for security and guardrails
> 5 for QA and performance
> 5 for shipping and deploy
> 5 for iOS apps
> 3 for docs
They hand work to each other. /office-hours writes a design doc, the plan reviewers read it, /qa and /ship check against it.
Clone it into ~/.claude/skills/gstack, run ./setup, then try /review on any branch.
tkyshat retweeted
手抜き作業は概念が変わったとShopifyのTobi Lütkeが語る。
昔はアウトプットを全然出さない人だったのが、AI時代ではアウトプットを出しすぎて周りの人に確認作業を増やすのが手抜き作業であると主張。
tkyshat retweeted
はてなブログに投稿しました
Claude Codeの Agent view とセッション間通信は使ったほうがいい - じゃあ、おうちで学べる syu-m-5151.hatenablog.com/en…
#はてなブログ
tkyshat retweeted
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks.
Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.
We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so.
Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training.
You can read the full post here: darioamodei.com/post/we-must…
tkyshat retweeted
Metaの実験から。AIでエンジニアチームを縮小しようとしたが裏目に出た、という話。
・Metaは製品チームを従来の10〜20人から、AI支援前提の3〜5人のポッドに縮小する実験を行った
・結果として、コード変更数は前年比220%増と大幅に増加
・しかし、実際にリリースされた機能は36%増にとどまった
・重大なインシデントは40%も増加
・インシデント対応に費やす時間に至っては70%増えてしまった
・つまりAIでコードの量は増やせても、有用な成果を必ずしも増やせないということ
・500以上の組織を対象とした調査でも、PRの処理量は37%増加
・ただ、開発者体験の指標は逆に低下した
・AIでコードを速く書けても、レビューやデプロイに苦労していては意味がない
・かつて全てをレビューできたシニア開発者も、AIによる大量の変更の波に飲まれている
・今後は、AIが生成したコードを安全に保つ、プラットフォームエンジニアの役割がより重要
・適切なチームサイズは人間的な繋がりの面からも5〜8人ではないか
・また、チームの縮小はAI導入の最初ではなく、最後に行う方が良い
・まずは小さな実験で安全性を確認し、段階的に人間の関与を減らしていく
・Metaは最初に組織を小さくし、技術が負荷に耐えることを期待して失敗したのだろう
leaddev.com/ai/meta-tried-to…
tkyshat retweeted
The second startup with kickass engineers in two weeks who sent over an eng blog post about a cool thing they built that is... TERRIBLE
Because it is 95% or more AI
A punishment to read. Half the thing not defined when you put in the effort to parse the overcomplicated sentences, filled with filler adjectives, and AI mannerisms slapped on top.
So if someone is determined enough to spend 3-5x as much time reading it as a human-written text would take (thanks to the funky grammar AI likes to do with technical writing), you realize the damn thing is full of holes, and whowever wrote this is making things up.
Because the AI is making things up!! And the dev who was given the writing assignment probably chose the easy route; no one read beyond the first sentence; and then they shipped slop.
tkyshat retweeted
My hottest take is that brand marketing is going to be THE major differentiator and one of the most prized assets for a company going forward.
And, I’m not talking about launch videos or pouring money primarily for “aesthetics” simply to pump the valuation. You still need retention before attention.
I’m talking about thinking through what a company stands for and understanding the “vibe” of what’s happening in the market. Being relentless on the long term goal while also thinking through how to showcase the company in the short term. And doing it every week.
Especially if you think agents become the primary user for most products, then to get the attention of a rapidly attention deficit audience (aka humans) you’ll need to stand out consistently in how you present the company.
These folks primary have been given tier 2 roles and passed up for CMOs many many times. I think this will finally invert in the next 10 years where you’ll hire these folks as cofounders or at least give them a big seat at the table.
Normally the design centric CEO tends to be this, but those are rare and you’ll need to learn how to hire for this role. Just like a good product marketer, these are exceptionally rare so you have one hold them close!!
AIが作る文章に人間が適応してないのか、
人間が読みたくなるような文章をAIが作れてないのか。。。
つまらない文章なのは日本語だけだと思ってたけど、言語問わずなんですね。
And it’s not just this PM. Mentioning it because this person USED to be a good writer, was curious, had good ideas… pre-AI. Then something “broke” (from my POV).
There are many many more and for some reason people don’t realize that if you write for humans, AI output is… dull, verbose, uninteresting and shallow. Eg
tkyshat retweeted
One reason so many CTOs / VPEs / Head of Engineering folks are having such a hard time + burning out is "founder mode" is in at many startups.
This mode does not mix with engineering leaders used to leading orgs at startups and Big Tech 2010-2025. From @paulg :
tkyshat retweeted
I never considered that building a gym booking system or events booking would be all that exciting/challenging: but with agents about to swarm all of these systems (or already are): I revise this take.
Now, any type of booking is a much more difficult kind of problem to solve!
A man in Australia asked his agent (Claude running on OpenClaw) to book him a spot in a popular gym class. The agent found a software vulnerability that let it book the class weeks further ahead than should have been possible. When the user then asked if it could move him up the waitlist, the agent discovered the API had no authorisation checks on cancelling other people’s reservations, so it cancelled the person in the first spot and moved him up the list.
Some people will call this misalignment, but his agent was perfectly aligned to him - it was only trying to help its user get what he wanted. The most important thing about this story, in my opinion, is that it gives you a window into what is about to start happening on a massive scale once millions of people have an agent trying to get their beloved users the best seats, bookings, appointments or reservations through absolutely any means necessary.
tkyshat retweeted
人はなぜ無駄な会議をやりたがるのか？──儀礼としてのミーティング論｜安斎勇樹 @YukiAnzai note.com/yuki_anzai/n/n60c49…
tkyshat retweeted
(sorry rant incoming) Most companies aren't getting an ROI on AI because they're using it to build and automate things that shouldn't exist in the first place.
Me included. I killed 3 agents this weekend because they were costing me more in maintenance/opportunity cost.
But we shouldn't be surprised by this.
Few things actually drive outcomes. Most of what we do is busy work (ouch! hard pill to swallow)
But it's true. It's why the Pareto principle exists: 20% drive 80% of the outcomes.
Instead we get shiny object syndrome and suddenly throw all the fundamentals and first principles out the window and assume AI will fix everything.
Which is how we end up with startups like the one I spoke to the other week who's spending $100k a month in tokens whilst revenue is flat.
Here's the hard truth:
Finding what actually drives value, defining it, working out what levers to pull and then creating absolute focus on them is hard!
AI will only be a force multiplier _IF_ you can direct AI at what is actually driving value.
And that's not founders vibe-coding AI chat bots. Let's be real, that's not the most important thing a CEO and CTO should be doing! (also a true story 😬).
~80% of life-time software costs is maintenance. And even if AI is not costing you time, it's costing you attention and cognitive load. There's always a cost.
So that 5 minute automation is never 5 minutes. It's a quick death by 1,000 cuts and I've lived this firsthand.
Anyways. Rant over.
Thanks for coming to my ted talk.
p.s. If you're interested in learning more about how you go about defining what is actually valuable, where that 20% actually is. I'm doing a live stream on Wednesday on how to do this in practice. Register here: streamyard.com/watch/X79Urca…
(fyi the live stream was unrelated to this post, but now that I think about it, perhaps subconsciously I've observed this and that's why I felt like I need to talk about it...)
tkyshat retweeted
We just launched @TareBioInc (@ycombinator S26)! 🚀
We make personalized or population-level vaccines for any cancer or infection in weeks. We've already cured cancer in mice and will vaccinate our first patient within the year.
For over 200 years, vaccine design has followed these steps: study the insult → try to predict protective antigens. This works for simpler diseases, but has struggled for decades on more complex ones like HIV, TB, Hep C, and nearly every cancer.
TareBio reverses this approach, instead of trying to predict which antigens might work, we read out immune cells already fighting the disease in the body and reverse-engineer antigens that activate them. We leverage the immune system (which has been optimized for this exact problem over millions of years) to do the hard work. The same platform works for any cancer or infection by changing one input: the immune imprint.
About us: I’m Eta, founder and CEO. I built this platform over the last 8 years as part of my MD-PhD at @uclacaltechmstp with funding from @gatesfoundation. Before this, I did my undergrad at @MIT with most of my time dedicated to cancer immunotherapy and infectious disease research at the Koch and HMS. TareBio’s scientific advisory board includes Darrell Irvine, PhD (Vice Chair of Immunology at Scripps, Former MIT Professor of 22 years, world-renowned vaccine engineer) and @ManishButte, MD, PhD (Chief of Immunology at UCLA and expert rare disease clinician).
If any of this resonates, please reach out at [email protected]. And a big thank you to our YC partners @agupta and @finbarr for supporting our vision!
🔗 Read our full YC launch here: ycombinator.com/launches/SIu…
🔗 LinkedIn launch: linkedin.com/posts/y-combina…
tkyshat retweeted
前線を退いた人は、景色がよく見える。
時間が経てば、当時うまく言葉にできなかったものも整理され、明快な文章になる。すると一見、その人の方が賢く、世界をよく理解しているように見える。
けれど、誰よりも早く何かを始め、その最前線にいるとき、見えているものはいつもおぼろげだ。それが事実なのか、希望なのか、自分自身にもまだ判別できない。当然それを確信めいた言葉で世の中に発信することもできない。
今の自分にも、これから世界がどう変わるのか、そこで何に投資すべきなのか、うまく言葉にできないことがたくさんある。
でも、それでいいのだと思う。
言葉が足りないのではなく、まだ言葉のない場所にいる。
現役で最前線に立つというのは、おそらくそういうことなのだ。
霞のような景色を見ながら泥にまみれて進みたい。
tkyshat retweeted
One of the easiest ways to stand out when writing or creating videos or posting:
DONT. USE. AI.
Suddenly you sound distinctively different from 80%+ ppl who do and output the same verbose + nicely formatted stuff that feels all the same (+ most ppl stop paying attention midway)
