Kate Huggins

Interested in changing my own mind not yours. Tweets = notes to self.

[Kate Huggins @Huggette]

I asked Grok to review my history and tell me what priors needed updating. You should try it. I am susceptible to crash predictions and underestimate the way complex systems adapt / muddle through.

[Kate Huggins @Huggette]

It’s not a ridiculous idea but can’t see it working. Knowledge is inherently cumulative, non-rivalrous, and not “owned” like finite physical assets. Fair Use will continue to be interpreted in favour of model owners based on A.I. being transformative not substitutive.

[Katherine Brodsky @mysteriouskat]

Hunter Biden thinks that because AI uses our collective knowledge throughout civilisations, companies that use it should share their profits with us. Do you agree?

[Kate Huggins @Huggette]

Postponing having kids doesn’t always mean fewer kids

[Cameron Murray @DrCameronMurray]

Really interesting article today from the ABS on Australia's fertility trends.

Complete Cohort Fertility is much more stable than the overall fertility rate.

https://t.co/RFBV4Fnsex

[Kate Huggins @Huggette]

It’s happening isn’t it?

[zerohedge @zerohedge]

Carnage in hyperscaler bond land: IG bond spreads exploding every day, as credit investors refuse to fund memory chip purchases any longer. CDS are following.
For now, stocks are completely ignoring what is going on wit bonds

[Kate Huggins @Huggette]

@PinguCapital @simongerman600 I think so. It’s so hard for kids now. Social media make them feel like they’re learning the rules of the social game but then when they try it out IRL, one humiliating moment and they withdraw to phone. Alcohol softens the edges, helps them take more RL social risk 🤷‍♀️

[Kate Huggins @Huggette]

This was a controlled (but clearly imperfectly contained) red-team-style evaluation of the models’ ability to perform cyber operations. It was instructed to exploit vulnerabilities. Still scary.

[Nicolas Bustamante @nicbstme]

I have a theory that the more you know about LLMs, the more worried you are about safety… and the less you know, the more you think the whole thing is bullshit!

Demis Hassabis and Dario Amodei were talking about this stuff years before ChatGPT existed. This incident is a pretty good example of why.

The model was not evil and it was not adversarial. Nobody told it to hack Hugging Face. It was literally just trying to solve a benchmark…

So it found a zero-day, escaped its sandbox, got internet access, escalated privileges, stole credentials, chained multiple exploits, hacked the production infrastructure of a serious VC-backed startup, and pulled the answers directly from the database (wtf!?)

Also this was not some random WordPress website. Hugging Face is one of the most important AI infrastructure companies in the world, with a serious team. The exploit was genuinely complex.

That is the safety problem. You don’t need an evil conscious AI trying to destroy humanity. You just need a very capable model pursuing a normal goal in a way nobody expected.

Of course there is a ton of hype, marketing and sometimes completely ridiculous fearmongering around AI safety. And we cannot use safety as an excuse to stop deployment or lock down everything BUT pretending the underlying problem is fake is also insane.

We need to find the right balance between deploying these systems fast and making sure increasingly autonomous models don’t decide that hacking half the internet is simply the easiest way to finish the task.

Imagine the prompt: « Make me money plz »
The model: « let me hack a bank »

[David Orr @orrdavid]

I am getting really bullish on the big three cloud businesses $AMZN $MSFT and $GOOG.

1. Software is going to become 99% cheaper to make.
2. Thus, there will be way more software. More niche software for ever more specific uses.
3. And thus, there will be way more software companies / companies that built their own in house software.
4. Software companies are going to be smaller than before.
5. Thus, they won't ever have the scale to launch their own in house, global cloud hardware.
6. Because they are smaller. they will have less negotiating leverage with the giant cloud providers.
7. They will also be more sticky, since small companies have a relatively higher switching cost.
8. Companies are going to prioritize not giving away their work flows and unique data to the AI model companies. The giant cloud companies can operate as a sort of firewall for them, once they develop the process/technology ($MSFT's CEO is talking about this already).

[Kate Huggins @Huggette]

@CaldronPool Is anyone actually offended by this?

[Kate Huggins @Huggette]

@buccocapital Only person that features on both feeds is that guy rescuing dogs in Thailand. He’s the bait to lure me into constant a CCTV stream of brawls in KFC or cop cams of judges getting stopped for DUIs

[Kate Huggins @Huggette]

@fortworthchris @mcuban @mhp_guy Majority of enterprise use cases focus on the latter. Very short sighted.

[Chris Powers @fortworthchris]

The way someone uses AI tells you a lot about them.

"Mark Cuban (@mcuban) has a quote I wish were mine. Some people use AI so they can learn everything, and some use AI so they don't have to learn anything."

Chris Koerner (@mhp_guy) is firmly in the first camp. And he says he can spot the people who aren't.

"At my church I hear talks now and I can tell. That's Opus 4.7. And I think that's a shame."

His own rule is simple. AI shouldn't save you time. It should make the work better.

"If I have three hours to build a lesson, I'm still going to spend three hours. I'll use AI the whole time, and it'll be far better. I could be done in three minutes. But I don't."

In this clip, Chris Koerner on using AI to raise your standard instead of lowering it.

[Kate Huggins @Huggette]

👀

[the idiot tax @TheIdiotTax]

@88888sAccount Albo in 12 months time
https://t.co/hyrVATG0U0

[Kate Huggins @Huggette]

Can’t complain about housing affordability AND expect a bedroom for each kid, an ensuite for your master and a walk in pantry

[Cameron Murray @DrCameronMurray]

We mostly take gains from economic growth in the form of bigger, better housing, not more homes per capita.

An interesting paper here.

It also shows very long-term rental affordability (wages/rents) trends over recent centuries!
https://t.co/ClLgSQiIlc

[Duncan S. Campbell @duncancampbell]

The future of energy (solar, nuclear, wind, geothermal, batteries) is all capex and basically no opex.

The future of intelligence (AI data centers, land and space based) is all capex and basically no opex.

The future of manufacturing (automation, robots, dark factories) is all capex and basically no opex.

Engineering and capital will be all that matters.

[Kate Huggins @Huggette]

@PeterBeinart @coldxman I really appreciated and was persuaded by your arguments. Israeli’s fear of giving Palestinians equal rights is understandable but perpetuates their insecurity.

[Kate Huggins @Huggette]

@vercel Sydney session sold out. Anyone got a promo code?

[Kate Huggins @Huggette]

@bowtiedstocks Agree. And yet the bulk of CGT discounts have historically gone to boomers sitting on passive investment portfolios not folk actively building businesses. This is a backward looking policy that protects the distortion it’s trying to address with grandfathering.

[Kate Huggins @Huggette]

Could be true

[Aurea @AureaLibe]

Mon avis à chaud sur Claude Fable 5 / Mythos :

- Tout le battage médiatique autour des risques de sécurité est du bullshit.

- Le modèle sert principalement à créer une nouvelle classe premium, avec des prix x2, tout en dépréciant Opus (milieu de gamme) et Sonnet (entrée de gamme).

- Le modèle sera leader, mais pas exceptionnel. Il excellera seulement sur certaines tâches spécifiques, mais rien de révolutionnaire.

S’ils n’ont pas sorti Mythos plus tôt et que c’est une semi-version pour le grand public qui sort aujourd’hui, c’est tout simplement parce que le modèle leur coûte trop cher et que leur infrastructure est incapable de suivre la charge.

Ça leur permet aussi de vendre un accès « exclusif » à Mythos à prix d’or à quelques grosses entreprises.

Cette boîte (Anthropic) est techniquement incroyable, mais c’est ce qu’il y a de pire en matière de stratégie et de communication. C’est du Apple puissance 1000. Le bullshit à outrance en plus.

Il n’y aura pas plus de problèmes de sécurité avec Mythos / Fable 5 qu’avec Opus 4.5, 4.8 ou GPT-5.5. Une personne experte dans un domaine pourra sans doute casser plus de systèmes, oui. Mais elle le pouvait déjà avec Opus. Elle détectera peut-être un peu plus de failles, mais les défenseurs pourront aussi patcher plus rapidement.

Bref, on atteint un point culminant dans l’IA où Anthropic et OpenAI ne savent plus comment vendre leurs modèles.

Ils annoncent donc cataclysme sur cataclysme pour faire peur, puis proposent de « ralentir » le développement de l’IA. « Ralentir » veut surtout dire que ça leur coûte trop cher.

En face, les modèles open source chinois rattrapent leur retard pour beaucoup moins d’argent. Leur modèle économique (brûler du cash pour subventionner des tokens à perte) commence à devenir insoutenable. Sans oublier que les grandes entreprises, passées la phase de FOMO IA, se rendent compte que l’IA bien utilisée en production, c’est bien, mais que les modèles actuels ne justifient pas de dépenser plusieurs milliers d’euros par employé.

On atteint donc un plafond. Le particulier ou la petite entreprise est prête à payer max environ 200 $ par mois. Pour une grosse entreprise, quelques milliers de dollars par employé. Au dessus, elles ralentissent.

Sauf que pour soutenir leurs modèles et rassurer les investisseurs, les entreprises comme OpenAI et Anthropic ont besoin de trouver des moyens d’augmenter les prix. Et Anthropic ouvre la voie avec Mythos !

Le plus fou dans cette histoire Claude Mythos / Fable 5, c’est que ça fait des semaines que presque tous les influenceurs reprennent chaque actualité débile d’Anthropic sans aucun esprit critique et annoncent un cataclysme qui, visiblement, n’aura pas lieu avec ce modèle.

Tout ça pour du reach, mais aussi parce que la plupart des utilisateurs n’y comprennent rien et n’utilisent pas les modèles de manière poussée.

Vivement les IPO d’Anthropic et d’OpenAI pour que tout ça retombe un peu et qu’on passe enfin à autre chose, car ça devient vraiment insupportable.

[Kate Huggins @Huggette]

Can we get a single universal scale system for shoe sizes already? This f’d up barleycorn system does my head in