HNHacker News
TopNewBestAskShowJobs

stared

14,247 karma · joined September 2, 2012

A curious being, doctor of sorcery. Posts, projects & resume at: https://p.migdal.pl/

Now: benchmarking AI at: https://quesma.com/blog/

Previously: co-founder & CTO at https://quantumflytrap.com/

submissionscomments

Opus 5.5 causing an identity crisis is real

kondz.io·20 pts·stared·
41

When do people experience mental health conditions for the first time?

ourworldindata.org·2 pts·stared·
0

I gave Opus 5.5 one prompt and six hours to visualize Invisible Cities

quesma.com·418 pts·stared·
199

Using real shadows to make a video game

instagram.com·3 pts·stared·
1

Show HN: Navier-Stokes Equation Explained Colorfully

p.migdal.pl·1 pts·stared·
0

Is GLM-5.3-Flash Mythos-Level at Cyber?

generality.org·2 pts·stared·
0

Show HN: Truth, druid, dryad, tar and dendrite share the same PIE root as tree

p.migdal.pl·1 pts·stared·
0

Agent session transcripts are precious, keep them

quesma.com·2 pts·stared·
0

DotFly – a fruit fly's brain, simulated in C#

kokosa.dev·1 pts·stared·
0

Claude couldn't hack OpenAI. Then Anthropic shipped Opus 5

thenewstack.io·13 pts·stared·
1

GPT-6 Astra solves puzzles

quesma.com·2 pts·stared·
0

Knowledge vs. wisdom: asking AI "What mushroom is that?"

quesma.com·2 pts·stared·
0

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

quesma.com·286 pts·stared·
139

I threw a ton of various logic puzzles at Astra and it solved of them

twitter.com·1 pts·stared·
0

GPT-6 Astra has autonomously completed Portal

twitter.com·8 pts·stared·
0

Does your landing page look generated?

slop-detect.com·3 pts·stared·
3

A connectomics milestone: Mapping the complete male fruit fly brain

research.google·2 pts·stared·
0

Show HN: Genetic Distance Map

p.migdal.pl·2 pts·stared·
1

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

quesma.com·3 pts·stared·
1

Beef production in Brazil has been the largest driver of global deforestation

ourworldindata.org·5 pts·stared·
0

Mushroom hunting with LLMs: what can go wrong?

quesma.com·56 pts·stared·
76

Witcher 3's DLC Won't Include Russian Voice Acting

thegamer.com·5 pts·stared·
3

Forget the Pelican, It's Weevil-Time Benchmaxxing-Proof SVG and Vision Test

reddit.com·1 pts·stared·
0

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

quesma.com·3 pts·stared·
2

Benchmarking Qwen3.8 27B quantizations: 4-bit holds up, 1-bit collapses

quesma.com·15 pts·stared·
2

If Eastern Europeans reported the world like the world reports Eastern Europe

badbaltic.com·4 pts·stared·
0

Gemini 3.7 Flash, Grok 4.6, GLM-5.3 and DeepSeek V4 Pro joined the frontier

quesma.com·2 pts·stared·
0

NanoGPT Speedrun Frontier

primeintellect.ai·140 pts·stared·
39

Gemini 3.7 Flash, Grok 4.6, GLM-5.3 and DeepSeek V4 Pro joined the frontier

quesma.com·5 pts·stared·
0

Pander Score: How much do AI models mirror what users believe?

sophronresearch.org·2 pts·stared·
0
Page 1 of 29Next →