dnouri @dnouri

~ХАЙСЕ ШАЙСЕ~ danielnouri.org Sweden Joined April 2009

Tweets

6K
Followers

2K
Following

828
Likes

4K

Matei Zaharia @matei_zaharia

2 days ago

3) Harnesses make a huge difference in cost-performance. The very simple Pi harness (@badlogicgames) got the same success rate as harnesses from the LLM vendors with Opus and GPT 5.5, but at 2x less cost! Seems to be mainly due to smaller inputs to the LLM.

12 39 384 102K 233

View Details

dnouri @dnouri

3 days ago

Hello, Sofielund!

0 0 0 24 0

View Details

dnouri @dnouri

3 days ago

GitHub: github.com/dnouri/crimeci… App: crimecity.danielnouri.org

1 0 0 26 0

View Details

dnouri @dnouri

3 days ago

This might be the best way to find out when the last murder happened in your Malmö neighbourhood! We built an open source, public-data app for exploring 80,000+ Swedish police incident notices.

1 0 0 126 1

View Details

dnouri @dnouri

2 weeks ago

Apparently you can go pretty crazy styling the Pi coding agent Emacs frontend :)

1 0 1 156 1

View Details

Manolo De Los Santos @manolo_realengo

2 weeks ago

The US has seized $5.49 billion of Venezuela’s oil revenue in 2026, then offers just $150 million in earthquake aid while the UN estimates $6.7 billion in losses. It’s not relief, it’s extortion. At the same time, US has $30 billion of frozen assets while people are under rubble.

430 8K 16K 397K 935

View Details

dnouri @dnouri

2 weeks ago

GPT-5.5: > I might feel a bit irritated because the user said "Then write," which feels more directive compared to "ask first."

0 0 0 100 0

View Details

dnouri @dnouri

2 weeks ago

EU just needs to partner with China, that's all.

Andrew Curran @AndrewCurran_

2 weeks ago

Ominous. It would be seismic if the general release of GPT-5.6 turned out to be US only. It would also almost guarantee that Fable 5 would get the same treatment. I did expect this eventually, and have said so, but not in summer 2026. Not this soon. Everything is speeding up.

98 100 1K 231K 149

0 0 0 73 0

View Details

dnouri @dnouri

2 weeks ago

Relatedly, consider a GPT-5.6 that would have been released without the government scrutiny theatre. It's a measure of quality that the USG give you a stamp and not allow you to release your model because it's too good. OpenAI's marketing and USG scratching each other's back.

0 0 0 46 0

View Details

dnouri @dnouri

2 weeks ago

Given how quickly a GPT-5.6 happened, on par with Mythos, it's fair to assume that OpenAI had this type of model for some time, unreleased, and haven't just *caught up within 2 weeks*. Consumers having only access to tier 2 models across providers is likely not a new thing, thus.

Chubby♨️ @kimmonismus

2 weeks ago

It is absolutely crazy how the last two weeks have changed the entire future. It is unprecedented that access to "frontier" models was cut off,and presumably remains cut off forever. It feels like a watershed moment, as if access to the highest level of human intelligence had

140 133 2K 58K 150

1 0 0 175 0

View Details

𝕊𝕡𝕣𝕚𝕟𝕥𝕖𝕣 𝕻𝕣𝕖𝕤𝕤 @SprinterPress

3 weeks ago

The average life expectancy in Gaza has dropped from 75 to 35 years. This has never been recorded before in medical history. -Dr. Mads Gilbert

17 289 496 17K 46

View Details

Gary Bernhardt @garybernhardt

3 weeks ago

Both of these were on new subsystems. If you let it dump all of the slop in repeatedly, the slop grows superlinearly. It slops up a bunch of complicated interfaces to afford mocking, then in the next change it has to work around them to "preserve backwards compatibility".

7 14 529 56K 24

View Details

Z.ai @Zai_org

3 weeks ago

Introducing GLM-5.2: Frontier Intelligence, Open Weights - Significant improvements in coding and agentic tasks - Strong long-horizon capabilities with a 1M context window - Two levels of reasoning effort: GLM-5.2 (max) pushes the limits, while GLM-5.2 (high) strikes a strong balance between performance and token efficiency - MIT-licensed open weights - Same API pricing as GLM-5.1 Tech Blog: z.ai/blog/glm-5.2 Weights: huggingface.co/zai-org/GLM-5.2 API: docs.z.ai/guides/llm/glm… Coding Plan: z.ai/subscribe Chat: chat.z.ai

710 2K 13K 7.5M 5K

View Details

dnouri @dnouri

3 weeks ago

There's some brilliant observations and advice for day-to-day AI-assisted coding here.

Mitchell Hashimoto @mitchellh

4 weeks ago

My heuristic is that any diff an agent generates over ~1500 lines is too big and is indicative that the problem needs to be decomposed. This is my general pattern now for feature work: 1. Try to implement the whole feature, loosely guided. I call this the "draw the owl" prompt

97 233 4K 212K 3K

0 0 0 129 0

View Details

dnouri @dnouri

4 weeks ago

New blog: The rise and fall of Fable Wherein I discuss the latest drama around frontier LLMs, the role of Chinese models, introduce a new subagent extension for Pi, and use it to build falling-blocks game in Zig from one prompt. danielnouri.org/notes/2026/06/…

0 0 0 84 0

View Details

dnouri @dnouri

4 weeks ago

.@Fortnite really pulled a feat with the new season. Those sprites seem to appeal to the Steal the Brainrot “collect them all” mind so much that people I’m playing BR with no longer care about their health, weapons, nor winning. It’s all about the golden peanut sprite now. 🤦‍♂️

0 0 0 99 0

View Details

Jeremy Howard @jeremyphoward

4 weeks ago

I disagree with this decision and I don't like it. But also... HOW DID ANTHROPIC NOT SEE THIS COMING‽ It is *the* obvious response to "this is too dangerous for anyone except us to use", since that relies on a premise ("we are uniquely good") that almost no-one agrees with.

Anthropic @AnthropicAI

4 weeks ago

The US government, citing national security authorities, has issued an export control directive to suspend all access to Fable 5 and Mythos 5 by any foreign national, whether inside or outside the United States, including foreign national Anthropic employees. The net effect of

13K 26K 88K 92.5M 24K

237 209 3K 224K 217

View Details

dnouri @dnouri

4 weeks ago

@francoisfleuret Because the idea that China has a caste system is completely made up and your video does not show what the person posting it pretends it shows. But the idea evidently serves so well your narrative of the "enlightened West" versus brute yellow man. Here's ChatGPT for you:

1 0 1 157 0

View Details

antirez @antirez

a month ago

I find myself telling coding agents on different machines to take a log about the work the are doing, and to use different skill files for certain processes. It's time to start using Redis arrays with ARGREP I guess, to have all centralized. I'll share a skill file and a video.