Sibylle Tretera

Tech Leader, thinker, writer, leadership coach. Google I Pinterest I Getty I Omnicom. Expat living in the fast lane. Science and History lover.

[gabriel @gabriel1]

we're hiring! for onsites we fly people to sweden until our US visas are done

you'll enjoy working with us if you always have work top of mind, enjoy rational product dicussions, and want to build/sell/market something amazing

reach out through a mutual or otherwise dms!

[Boyan Slat @BoyanSlat]

Excited for @TheOceanCleanup to come to the Manila Bay region! Today we signed with the Department of Environment and Natural Resources to work together on stopping plastic pollution in the Pasig River before it reaches the ocean.

[derek guy @dieworkwear]

i genuinely hate how internet culture has become so normalized and widespread. i do not want a white house deputy chief of staff saying "it's white boy november" or a secretary of state tweeting that he's "locked in for this banger." grow up.

[Claude @claudeai]

Available on all paid plans.

Read more: https://t.co/nOrCwKwTwL

[AYi @AYi_AInotes]

Damn，Karpathy这条帖子直接把我过去半年的AI工作流全推翻了🤯

大家都在死等更强的模型，
死等更大的上下文窗口，

但Karpathy说，你们全搞错方向了，
现在AI最大的瓶颈，根本不是模型不够聪明，
是我们还在用文本这种最低带宽的方式，跟它沟通。

他推荐了一个所有人今天就能用的trick，
在任何query的最后加一句：
"structure your response as HTML"
然后让Claude直接帮你打开，

出来的效果好到离谱，
不仅仅是��了点颜色和排版，
更像是你终于给AI打开了大脑里��片10车道的视觉超级高速公路，

同样的内容，HTML的阅读效率和理解深度，是Markdown的10倍以上，

这简直就是人机交互的真正下一代范式，因为人类的输入和输出偏好，天生就是完全不对称的，

输入最自然的是音频，说话比打字快4倍，思考也更连贯，
输出最擅长的是视觉，我们大脑1/3的皮层，全用来处理视觉信息，

而我们现在，却在用文本这种单车道的土路，双向跑所有的流量，

Karpathy画了一条清晰的演进路线：

原始文本 → Markdown → HTML → 交互式神经视频，

我们现在正站在Markdown到HTML的转折点上，

最令人兴奋的是，很多人说HTML费token，生成慢，

但你算一笔账就懂了，
多花2倍的token，换你10倍的阅读速度和理解深度，
这是全世界最划算的交易了吧哈哈，

可惜我们早就被省token的思维绑架了，却忘了人类的时间才是真正的稀缺资源，

还有一个更扎心的认知，
Markdown是给AI看的格式，
HTML是给人用的格式，

AI代理之间沟通，用Markdown甚至JSON都没问题，
但所有最终要给人类消费的东西，都应该切成HTML，
这才是最优的分工，

现在我已经把所有prompt的结尾，都加上了那行字，
做对比用并排表格，做分析用彩色标注，做原型用交互式滑块，
AI不再是给我甩一大段干巴巴的文字让我啃，
它直接给我造了一个可交互的视觉思考空间，

Karpathy说，人机的心智融合才刚刚开始，
我们根本不用等Neuralink那种脑机接口，
先把HTML用起来，就是当下能摘到的最大最甜的低垂果实🍒

#AI #Karpathy

[David Sirota @davidsirota]

Destroying the @InternetArchive's @WayBackMachine would be the equivalent of the burning of the Library of Alexandria - one of the worst losses of knowledge in history.

Media giants are now threatening to do this.

We can't let this happen.

Pass it on.

[Tuna @antea04]

I'm a data scientist @OurWorldinData and I need help from a botanist or someone local to Kyoto, Japan! 🌸

We present one of the world’s longest climate records: 1,200 years of peak cherry blossom dates in Kyoto.

The researcher who maintained it, Professor Yasuyuki Aono, sadly passed away last year.

[Anthropic @AnthropicAI]

Introducing Project Glasswing: an urgent initiative to help secure the world’s most critical software.

It’s powered by our newest frontier model, Claude Mythos Preview, which can find software vulnerabilities better than all but the most skilled humans.
https://t.co/NQ7IfEtYk7

[Norwich Cathedral @Nrw_Cathedral]

A second peregrine falcon egg has been laid on the Cathedral spire! 🥚🥚

Egg number two is thought to have been laid at about 8.50am today.

https://t.co/ghepPt4Uec

📷 @Hawkandowluk

[Sibylle Tretera @sib_tretera]

@_chenglou fantastic work @_chenglou - and thank you for sharing with the community! I’m working with clients and brands who are keen to optimise for AI - so this is very timely!

[Cheng Lou @_chenglou]

My dear front-end developers (and anyone who’s interested in the future of interfaces):

I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept):
Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

[Alex Finn @AlexFinn]

This is potentially the biggest news of the year

Google just released TurboQuant. An algorithm that makes LLM’s smaller and faster, without losing quality

Meaning that 16gb Mac Mini now can run INCREDIBLE AI models. Completely locally, free, and secure

This also means:

• Much larger context windows possible with way less slowdown and degradation
• You’ll be able to run high quality AI on your phone
• Speed and quality up. Prices down.

The people who made fun of you for buying a Mac Mini now have major egg on their face.

This pushes all of AI forward in a such a MASSIVE way

It can’t be stated enough: props to Google for releasing this for all. They could have gatekept it for themselves like I imagine a lot of other big AI labs would have. They didn’t. They decided to advance humanity.

2026 is going to be the biggest year in human history.

[Sibylle Tretera @sib_tretera]

I still miss @Hedone 's bread. Fantastic journey also - M&A lawyer -> food blogger -> chef -> @Michelin awarded -> baker. In case anyone knows what Mikael JONSSON is currently up to?
Always judge a restaurant by its bread basket https://t.co/MZ9SWJjysP via @ft

[Dan Neidle @DanNeidle]

I see some weird things but this takes the biscuit. A vulnerability in the Companies House website, that let anyone view the private dashboard of any one of the five million registered companies, see directors' personal details.

And modify them.

[Seb Johnson @SebJohnsonUK]

Be Matt Clifford and Alice Bentinck:

> Meet at McKinsey in 2009
> Realise the top minds in the country are wasting their entering finance and consulting instead of building tech
> Spend 4 years running EF as a non-profit trying to make it work
> Raise £8.5m in 2015 to invest
> Back Magic Pony that gets acquired for $150m 18 months later.
> Instantly gain recognition in the industry
> Raise more money
> Back first unicorn
> Expand across the world
> Raise more money
> Undo Global expansion
> Double down on the US
> See revenue growth explode, time to raise shrink and valuations double
> Raise $200m and become a unicorn

In that time they've backed some generational founders including @Barney_H_Y, @jamesdacombe and @nikola_mrksic and MANY MORE.

Their portfolio companies are now valued at $15bn.

@Alicebentinck and @matthewclifford have transformed the ecosystem through sheer will.

And NONE of this includes all the work they've done outside @join_ef:

> Alice founded Code First Girls and has a book published
> Matt is Chair of ARIA, served as Adviser on AI to the PM after drafting an “AI Action Plan” for the UK and is now focused on making the UK rich again

Two of the greats and I couldn't be happier for them.

[TMZ @TMZ]

Daniel Radcliffe believes child actors need mandatory therapy. https://t.co/oesh4MAFKj

[Sibylle Tretera @sib_tretera]

@soigomaa @netflix how about it? great share @soigomaa

[Computer @AskPerplexity]

Perplexity Computer replaced $225K/yr in marketing tools in a single weekend.

We built an AI marketing agent that scans hourly, manages budgets, detects fatigue, and coordinates several campaigns end to end.

In one test run, it made 224 micro-optimizations to our ad stack.

[Andrej Karpathy @karpathy]

Three days ago I left autoresearch tuning nanochat for ~2 days on depth=12 model. It found ~20 changes that improved the validation loss. I tested these changes yesterday and all of them were additive and transferred to larger (depth=24) models. Stacking up all of these changes, today I measured that the leaderboard's "Time to GPT-2" drops from 2.02 hours to 1.80 hours (~11% improvement), this will be the new leaderboard entry. So yes, these are real improvements and they make an actual difference. I am mildly surprised that my very first naive attempt already worked this well on top of what I thought was already a fairly manually well-tuned project.

This is a first for me because I am very used to doing the iterative optimization of neural network training manually. You come up with ideas, you implement them, you check if they work (better validation loss), you come up with new ideas based on that, you read some papers for inspiration, etc etc. This is the bread and butter of what I do daily for 2 decades. Seeing the agent do this entire workflow end-to-end and all by itself as it worked through approx. 700 changes autonomously is wild. It really looked at the sequence of results of experiments and used that to plan the next ones. It's not novel, ground-breaking "research" (yet), but all the adjustments are "real", I didn't find them manually previously, and they stack up and actually improved nanochat. Among the bigger things e.g.:

- It noticed an oversight that my parameterless QKnorm didn't have a scaler multiplier attached, so my attention was too diffuse. The agent found multipliers to sharpen it, pointing to future work.
- It found that the Value Embeddings really like regularization and I wasn't applying any (oops).
- It found that my banded attention was too conservative (i forgot to tune it).
- It found that AdamW betas were all messed up.
- It tuned the weight decay schedule.
- It tuned the network initialization.

This is on top of all the tuning I've already done over a good amount of time. The exact commit is here, from this "round 1" of autoresearch. I am going to kick off "round 2", and in parallel I am looking at how multiple agents can collaborate to unlock parallelism.
https://t.co/WAz8aIztKT

All LLM frontier labs will do this. It's the final boss battle. It's a lot more complex at scale of course - you don't just have a single train. py file to tune. But doing it is "just engineering" and it's going to work. You spin up a swarm of agents, you have them collaborate to tune smaller models, you promote the most promising ideas to increasingly larger scales, and humans (optionally) contribute on the edges.

And more generally, *any* metric you care about that is reasonably efficient to evaluate (or that has more efficient proxy metrics such as training a smaller network) can be autoresearched by an agent swarm. It's worth thinking about whether your problem falls into this bucket too.

[Peter Steinberger 🦞 @steipete]

Models are getting so clever, it's almost scary.