The Deep Research problem — Benedict Evans OpenAI’s Deep Research is built for me, and I can’t use it. It’s another amazing demo, until it breaks. But it breaks in really interesting ways.

"LLMs are good at the things that computers are bad at, and bad at the things that computers are good at. [The] model is trying to work out what you probably mean (computers are really bad at this, but LLMs are good at it), " @benedictevans.bsky.social www.ben-evans.com/benedictevan...

18.02.2025 15:49 — 👍 0    🔁 0    💬 0    📌 0

"It’s easy for coffee to come out either too acidic or too bitter. If coffee is under-extracted, it’ll be acidic; if it’s over-extracted, it’ll be bitter. But these are both easy problems to solve. Add salt to coffee, and it’ll cancel out acidity. Add baking soda, and it’ll cancel out bitterness."

11.02.2025 17:08 — 👍 1    🔁 0    💬 0    📌 0

Microwaving your coffee changes it, and other lessons learned (while sipping espresso) in MIT’s coffee class This February marks the second year of the MIT class “Coffee Matters: Using the Breakerspace to Make the Perfect Cup.”

“It’s mass within time that matters most. If you weigh the beans after you grind them, and then you weigh the espresso that comes out, that’s got to be a factor of two within 30 seconds," says MIT's Jeffrey Grossman who teaches Coffee Matters. www.boston.com/news/the-bos...

11.02.2025 17:06 — 👍 0    🔁 0    💬 1    📌 0

I think Google's revenue model for Gemini is that a user must buy a subscription to turn off the cloying, simpering tone.

09.02.2025 14:28 — 👍 0    🔁 0    💬 0    📌 0

Try harder, Gemini 2.0.

09.02.2025 14:25 — 👍 0    🔁 0    💬 1    📌 0

The Man in the Midnight-Blue Six-Ply Italian-Milled Wool Suit A perfect suit, made by an expert tailor out of superlative fabric, would do nothing less than transform me.

"My martini was so excellent, I struggled not to cry." Such a nice piece on enjoying a suit.

www.theatlantic.com/magazine/arc...

09.02.2025 08:46 — 👍 0    🔁 0    💬 0    📌 0

A Portrait of Stalin in All His Murderous Contradictions (Published 2017) The second volume of Stephen Kotkin’s biography “Stalin” reveals the ideologue and the opportunist.

"In politics, spite generally plays the basest of roles," said Vladimir Lenin, near death, who criticized Joseph Stalin's handling of the Georgia Affair, which was a tussle over how much autonomy Georgia could have within the Soviet Union. Bio by Stephen Kotkin.
www.nytimes.com/2017/10/19/b...

04.02.2025 07:05 — 👍 1    🔁 0    💬 0    📌 0

When using @anthropic.com 's Claude, I don't know why #WebDevelopment coding examples from the prompts need to be wrapped in CSS when you are just interested in grabbing the barebones, for example the mechanics of using a toggle button in react. Compare to @deepseek.bsky.social 's example.

28.01.2025 20:23 — 👍 0    🔁 0    💬 0    📌 0

Instead of "artificial general intelligence", the CEO of Anthropic, Dario Amodei, writes in his essay, Machines Of Loving Grace, that he prefers "powerful AI" to describe the next generation of AI tools. But what does Claude say? darioamodei.com/machines-of-...

25.12.2024 22:58 — 👍 0    🔁 0    💬 0    📌 0

During his NeurIPS talk, Ilya Sutskever says "Pre-training as we know it will end", as "we've achieved peak data and there'll be no more" (Kylie Robison/The Verge)

Main Link | Techmeme Permalink

14.12.2024 01:41 — 👍 5    🔁 1    💬 0    📌 0

"I would say that it's definitely impossible to predict the future," says Ilya Sutskever at #neurips2024 #neurips. Trends he is watching for in #AI are agentic, reasons, understands and self- aware.

13.12.2024 22:42 — 👍 3    🔁 0    💬 0    📌 0

It's my last day at #NeurIPS2024. Some highlights were Building Industry-Grade NLP Applications For The Financial Domain by @bloomberg.com's Andy Liu and Active Training: Building Agentic Apps With Llama 3.2 And Llama Stack by Meta's Raghotham Murthy

13.12.2024 21:49 — 👍 0    🔁 0    💬 0    📌 0