Hussein Kanji

Build more tech. Eat more cake.

[Shakeel @ShakeelHashim]

Laura Gilbert — former Number 10 director of data science and incubator for AI, now senior director of AI at TBI — has published a *scathing* critique of how recent governments have treated science and tech policy, as seen from the inside.

[Rex Woodbury @rex_woodbury]

Last week we sent out our Q2 letter to Daybreak LPs.

In that letter, we tried to distill a concept we've been talking about internally: the Post-Agentic Founder. I.e., what founders succeed in an AI world?

Sharing here what we wrote to our LPs:

[Hussein Kanji @hkanji]

It’s so inexpensive to do stuff in the UK

[Mike Bird @Birdyword]

An amazing insight into the state of UK freelance journalism: apparently a plausible bung to try and get a puff piece placed in a major newspaper is... £100.

[Davit @DBuniatyan]

At Mistral’s NeurIPS 2023 afterparty, I saw Sarah taking out trash bags. I offered to help, and she said, “Nah, enjoy yourself. This is what it takes to support.”

That moment stuck with me. Very few investors would do this metaphorically, let alone literally.

Not surprised at all by her rapid success.

[David Klion @DavidKlion]

This piece has been accused of whitewashing Biden administration officials' complicity in Gaza. While officials who cooperated with the reporting surely intended that, the resulting article is packed with new details that utterly damn them. https://t.co/FCkiN3pnrh

[Augmental @augmentaltech]

MouthPad is a touchpad you drive with your tongue.

Now available across the US.

[Rowland Manthorpe @rowlsmanthorpe]

If you're wondering why splitting up DSIT is such a big deal, take a look at this internal document parcelling out its different responsibilities (which still doesn't answer every question about roles)

This is why this kind of thing can take anywhere from 6 months to 2 years

[Hugo Amsellem @HugoAmsellem]

SF is breaking spiritually with 3 philosophical crisis at once:

1) talent is existential - infinite options, no idea what's worth building anymore, so they either start smth or join a lab. every week i hear about another founder whose CTO left for anthropic.

2) the public is nihilistic - religion, neighborhood, family, stable couple: most sources of meaning were already stripped, the job was one of the last ones standing, and now the public feels AI is coming for the brain and robots for the body. i live near the openai office in SF, 100s of people demonstrate in front of it regularly.

3) founders are absurdist - building is a commodity, dozens of competitors appear the week you start, every model release kills hundreds of startups or forces a pivot, and the boulder rolls back down. il faut imaginer les founders heureux.

the deeper conundrum i feel: the industrial revolution freed our bodies from labor - we got fat, invented the gym (a synthetic physical activity, like fetching the ball for a dog), then GLP-1 to patch the symptom w/ unknown tail risk on a complex biological system.

the intelligence revolution is freeing our minds from labor and the same playbook is coming - a GLP-1 for the mind (i keep hearing nasal DMT sprays are close). same patch, probably worse tail risk.

and we're still biologically human, so the risk is society splitting into 3 estates: a noblesse replaying la cour du roi, bored and plotting zero-sum games all day (homo leveragus) - a clergé who logs off entirely (homo experiencus) - and a tiers-état dating AIs in second life (homo matrixus). i'm not sure how we "faire société" with that, and as a french i remember the pitchforks in the history books.

underneath all of it, civilization runs on two coordination mechanisms: it coordinates matter through engineering to build technology (we're great here), and it coordinates meaning through storytelling to build culture (we have a problem here). the intelligence revolution pours everything into the first and hollows out the second.

au travail!

[Shashank Joshi @shashj]

Pretty wild. “While operating in our sandboxed testing environment, our models spent a substantial amount of inference compute finding a way to obtain open Internet access, in pursuit of solving the evaluation problem. To gain access, the models identified and exploited a zero-day vulnerability … We consider this incident to be an unprecedented cyber incident, involving state-of-the-art cyber capabilities, and are responding accordingly.”

[Zaid Jilani @ZaidJilani]

Why are these people so angry all the time. They got a bunch of money and they seem angrier than the average dude who works at Waffle House.

[Rosemary Kelanic @RKelanic]

How is giving uranium enrichment to Saudi Arabia putting “America First?”

It isn’t.

Enrichment is a potential gateway to future weaponization. That’s why Iranian enrichment drove Trump and Bibi bonkers.

Why would the US want to help put Saudi Arabia on the path to the bomb???

[Nicholas Kristof @NickKristof]

President Trump reaches a 1-2-3 nuclear deal with Saudi Arabia that would allow the Saudis to enrich nuclear fuel with few safeguards, increasing the risk that it develops nuclear weapons. This is crazy. https://t.co/FWkpwi4AqY Neither Iran nor Saudi can be trusted with nukes.

[Nicholas Kristof @NickKristof]

The Trump administration's demolition of humanitarian aid led to the closure of 1700 sites for treating HIV/AIDS and to 77,000 children losing access to treatment, a study finds. https://t.co/RsGNCP9KqM And note that the cost of treating AIDS is just 12 cents a day per person.

[Nicholas Kristof @NickKristof]

.@bartov_omer warns that we may be incentivizing genocide by failing to provide accountability. He focuses on Gaza; I'd note that Sudan fits that model as well, for world leaders refuse to speak up about the UAE's role in financing mass atrocities by the RSF in Sudan.

[Ambassador Mike Huckabee @GovMikeHuckabee]

Would be easy to say NY Mayor is just an ignorant bigot, but he is not dumb. He's an evil anti-Semite who deliberately lies about serious things like "genocide" to stoke Jew hate. If anyone should be arrested, it's the Mayor of NY for incitement. NYC should be ashamed.

[Ishaan Tharoor @ishaantharoor]

Ah, it all makes much more sense now.

[Matt Harney @SaaSletter]

👀👀👀 *Excellent* thematic report from Morgan Stanley (170 pgs).

Select excerpts /1

[Hussein Kanji @hkanji]

@VishalGulati_ Exactly. That captures it way better.

[Hussein Kanji @hkanji]

Except you don’t know how the kitchen is cooking the chicken because the recipe (weights) is a black box. Is the recipe telling the kitchen to sneak in something for flavor? You only know when you do a test run of the dish.

[Xiaoyin Qu @quxiaoyin]

Many people still confuse Chinese model with Chinese inference:
- A model is like a recipe for a dish—e.g. orange chicken
- Claude keeps its recipe secret, charges the dish $50. You must eat in front of Dario and he will watch you the entire time. You can’t change the ingredients or cook it privately.
- Now China releases a free recipe that tastes just as good. You can use American chicken and cook it in your own kitchen (self-hosting), or have a U.S. cloud like AWS or Nebius cook it and pay $2-5.
- Chinese inference is DoorDashing the finished dish from Beijing. Your order and data go there, and you must eat Beijing chicken. It’s fair to worry about the safety of their ingredients.
- Regulate the kitchen, ingredients, and data. Don’t ban the recipe.

[anna-sofia @annasofialesiv]

very well put and profound, explains why better articulated futures architect the present and therefore why compounding is still misunderstood