Advertisement · 728 × 90

Posts by Daniel Bachler

Preview
Just Talk To It - the no-bs Way of Agentic Engineering | Peter Steinberger A practical guide to working with AI coding agents without the hype.

Great overview of current best practices for agentic coding tools. Recommended reading if you write code. steipete.me/posts/just-t...

6 months ago 4 0 0 0

Things I find surprising about AI in Sept 2025:
- lots of progress on coding agents, little on game-playing
- really great writing is still beyond LLMs, no real progress since 4o
- AI generated videos are really convincing now, but fake AI content is not flooding my timeline yet

6 months ago 4 0 0 0

Excellent thread on AI by my colleague Esteban - recommended!

8 months ago 2 0 0 0

h/t @eortizospina.bsky.social for spotting this

9 months ago 2 0 0 0
Preview
Evaluating the impact of two decades of USAID interventions and projecting the effects of defunding on mortality up to 2030: a retrospective impact evaluation and forecasting analysis USAID funding has significantly contributed to the reduction in adult and child mortality across low-income and middle-income countries over the past two decades. Our estimates show that, unless the a...

Forecasts suggest current steep funding cuts to USAID funding could lead to over 14 million extra deaths (range: 8.5–20 million) by 2030 - including 4.5 million children under 5 (3.1–5.9 million).

From a new paper by Daniella Medeiros Cavalcanti et al: www.thelancet.com/journals/lan...

9 months ago 4 3 2 0
Post image

Childhood leukemia used to be a death sentence for the vast majority of patients. Now, most children in rich countries are cured.

Cool article from @scientificdiscovery.dev for @ourworldindata.org on crucial progress I was completely unaware of: ourworldindata.org/childhood-le...

10 months ago 11 1 0 0

So far we’ve been lucky that LLM economics results in almost perfectly egalitarian access to capabilities. It could’ve gone differently - where the richest enjoy far better systems or one lab holds a monopoly. Here’s hoping this stays true.

10 months ago 0 0 0 0

Really nice to see these features in prod! Kudos to my colleague @sophiamersmann.bsky.social for building all of this! 🔥

10 months ago 9 0 0 0

According to SimilarWeb, chatgpt.com had 5.1 billion visits last April (13% increase relative to March), and about 55% of the visitors were younger than 35.

For comparison: Wikipedia.org had 3.7 billion visits (down 6%), with ~45% under 35

10 months ago 6 2 1 0
Advertisement
Preview
Palm Oil Explore palm oil production across the world and its impacts on the environment.

Palm oil is not as bad as it is often made out to be - it is extremely efficient regarding land use. See ourworldindata.org/palm-oil

11 months ago 8 0 1 0
Preview
Palm Oil Explore palm oil production across the world and its impacts on the environment.

Palm oil is not as bad as it is often perceived to be - yields per land area are extremely high. See ourworldindata.org/palm-oil

11 months ago 3 0 0 0

“…today’s antivirals are so effective that if taken during pregnancy, the risk of passing HIV to the infant is minimal. Without these drugs, around a quarter of babies born to HIV-positive mothers would also contract the virus – and half of them would die before the age of two.”

1 year ago 30 8 1 0

Leopold Aschenbrenner and Daniel Kokotajlo have both argued that China must not reach AGI first. Have either of them given any update on how they think about US labs reaching AGI while Trump is in the White House?

1 year ago 0 0 0 0

Yeah that is an interesting theory - but then I’m still surprised that there isn’t more experimentation in that space. Not that I mind of course, I just find it a bit counterintuitive.

1 year ago 0 0 0 0

Maybe but LLMs are pretty cheap nowadays and most spam is really terrible. I’d guess that even just using a cheap, small model could improve the rate of replies a ton.

1 year ago 0 0 1 0

Something pretty strange about where we are with LLMs is how bad the spam in my email inbox still is. I would have guessed we would routinely see super convincing, personalized spam by now.

1 year ago 1 0 0 0
Post image

The way the other nations exports are calculated in Trumps "tariff" chart is so incredibly dumb

1 year ago 1 0 0 0
Post image

Someone on x came up with this which seems really dumb and therefore plausible

1 year ago 1 0 0 0

Tell me your team is entirely comprised of over-confident twenty-somethings without telling me your team is entirely comprised of over-confident twenty-somethings

1 year ago 257 43 20 1
Advertisement

I find many non-fiction books are more verbose than I would like. I'd often prefer a distilled version with a page count of roughly a quarter. I'd be willing to pay more for the distilled version than the verbose one. Could this be a new revenue stream for writers in the age of AI?

1 year ago 0 0 0 0

"Any grants, contracts, or collaborations involving mRNA vaccines be reported up the chain to Health and Human Services Secretary Robert F. Kennedy Jr.’s office and the White House" - Lysenkoism at work in the new NIH.

1 year ago 77 49 6 5

Have you tried chatgpt deep research on it? It can be pretty good at stitching together sources

1 year ago 1 0 1 0
Homepage of Works in Progress magazine, with articles: Steam networks, King of fruits, Fertility on demand, The rise and fall of the Hanseatic League, The prehistoric psychopath, The failure of the land value tax, Chinese towers and American blocks.

Homepage of Works in Progress magazine, with articles: Steam networks, King of fruits, Fertility on demand, The rise and fall of the Hanseatic League, The prehistoric psychopath, The failure of the land value tax, Chinese towers and American blocks.

Our latest issue of Works in Progress dropped today!

- The steam networks of NYC
- Prehistoric violence
- Urbanism with Chinese characteristics
- Extending the fertility window
- The Hanseatic League's rise and fall
- The pineapple: the king of fruits
- The land value tax

worksinprogress.co

1 year ago 50 9 2 2
Post image

Most OECD countries fail to allocate even 0.15% of their national income to aid the world’s poorest countries, despite a commitment to do so. That's just $1 out of every $700.

Only Luxembourg, Sweden, and Norway meet this minimal UN target. New cuts will deepen this shortfall.

1 year ago 10 6 0 1
Preview
For many of us, it doesn’t cost much to improve someone’s life, and we can do much more of it Most countries spend less than 1% of their national income on foreign aid; even small increases could make a big difference.

My latest article on Our World in Data, looking at foreign aid and the difference we can make with personal donations:

ourworldindata.org/foreign-aid-...

1 year ago 88 28 1 1
Preview
The number of children in South Korea has fallen by 60% since its peak South Korea is undergoing one of the world’s most rapid demographic transitions. Fertility rates — the number of children a woman has over her lifetime — have fallen rapidly over the last 50 years, an...

South Korea’s under-15 population has halved in two generations. I knew the population was aging quickly, and it’s no surprise that means fewer children—but the absolute numbers are striking.

It's easy to look at fertility rates and forget about absolute numbers. That's a lot fewer kids around.

1 year ago 9 3 0 0

I think this is right. There is another interesting dynamic that for AI labs it would be very useful if academic research were freely available and so you have a powerful new force against established, paywalled scientific publishing.

1 year ago 1 1 0 0

THIS IS HUGE! A cancer vaccine developed by Yale University and DFCI shows MAJOR promise. Results of an early-phase trial reveal ALL patients with ADVANCED stage kidney cancer had successful anti-cancer immune responses AND have remained CANCER-FREE approximately THREE years after treatment. 🧪🧵⬇️

1 year ago 11261 3373 218 236

Let me know how it goes for you!

1 year ago 1 0 0 0
Advertisement

OpenAI plus users now have access to the "Deep Research" feature. I just tried it for the first time and asked it for a report on a niche programming topic (comparing computation expressions in F# vs Algebraic Effects in OCaml) and the result was better than any blog post I know of on the topic.

1 year ago 6 0 1 0