Pinned Tweet
11 taun ngebacot asuw...
Do you remember when you joined Twitter? I do! #MyTwitterAnniversary
memamg semakin diliat AA ini smakin aneh weighted benchmarknya
Fuck you, @ArtificialAnlys
Your new index says Claude Haiku 5.5 is smarter than DeepSeek V4.1 Flash 1017 and GLM-5.3.
Haiku is Anthropic's small cheap model. DeepSeek 1017 is the model Cursor put on its cost frontier this month, and Cursor says no model scores higher at those prices. I run it every day on my 4x 3090. I dont believe for a second that Haiku beats it.
Look at how they got there. 60% of the index now runs on private tests nobody outside can see. GDPval-AA, SciCode and the long context test are gone. In comes a private coding benchmark from Cognition. And the open weights gap goes from 3 points to 12 in 1 release, 4x. Released open models are now 26 points behind the top.
1 day before v5, on the old version, they gave Ling 3.1 Flash 1019 a 44. Now the best released open model gets 32.
The weights didnt change. The tests did, and you cant see the tests.
This is a bought benchmark. A leaderboard where you cant check a single question, that widens the gap 4x overnight and puts Haiku above DeepSeek, is marketing for closed labs. Its not measurement. I hope at least you were paid enough money to publish such garbage. Liars.
Publish the questions or stop calling it an index.
Raving linenya jele bgt anj
this is f*cking insane...
a guy reverse-engineered his Ray-Ban Meta glasses with Opus 5.5 and turned them into JARVIS for the race track
it reads the track in real time, draws the racing line on the asphalt and tells him when to brake, turn in and counter-steer
his lap: 1:52.4 → 1:47.1. minus 5.3 seconds in one session
here's how it works step by step:
step 1 → the glasses stream what he sees to his phone, 30 frames a second
step 2 → a vision model finds the asphalt, the curbs and the next corner in every frame, even with his head shaking
step 3 → Opus 5.5 plans the racing line: outside → apex → outside, and paints it right on the road
step 4 → green = full throttle. red = braking zone. orange = you're in the corner
step 5 → before every turn: "BRAKE · 80 m", countdown, then "turn in", then "apex"
step 6 → when the rear slides, it shows the drift angle and which way to counter-steer
step 7 → the voice in his ear: "brake… now", "hold the apex", "full throttle"
step 8 → a mini-map in the corner compares his line with the ideal one and shows the lap delta live
under the hood:
→ no cloud, everything runs on the phone
→ if it loses the road, the line fades out instead of lying to you
→ every lap is saved, so it learns where you lose time
F1 teams pay millions for a race engineer. he made one out of sunglasses
Readers added context they thought people might want to know
Ray-Ban Meta glasses have no display/AR for road overlays; video is rendered visualization. Poster's follow-up shares hypothetical prompt using ~5 fps video call to computer, not claimed 30 fps on-phone real-time.
x.com/N01ennn/status…
meta.com/ai-glasses/ray…
vrarwiki.com/wiki/Ray-Ban_M…
sad
Looks like this is going to be a week of Apple press releases, starting today with the new OLED iPad Mini
- 8.5" OLED display
- A20 Pro chip
- "Acoustic display" instead of speaker grilles
- IP67 rating
- N1 and C2 chips
- Starting at $699 (+$200 from last gen)
My take: WAY better screen, WAY better chip, so a very welcome refresh. Very curious to try this audio system. but also I can't find any mention of 120Hz or ProMotion on the website, so.... 60Hz?
apple.com/newsroom/2026/10/a…






























