David Grenwood (@DavGrew) header
David Grenwood (@DavGrew) profile picture

AI news with receipts. tiny worlds Claude builds in one HTML file. no hype, just what actually works

Joined October 2026
Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)Photo by David Grenwood (@DavGrew)
David Grenwood retweeted
28
13
158
16,335
David Grenwood retweeted
Gemini: off. Search: on. AI Overviews: still included. That's the catch in Google's new settings for K-12 student accounts. Google says Gemini, Gemini in Classroom and Google Search will be off by default for students, rolling out in the coming months. School admins decide what to turn on, one service at a time. So a school can give students Search without Gemini. What Google hasn't offered: Search without AI Overviews and AI Mode. Teachers and school IT folks, which would you pick: Search with AI Overviews, or no Search?
22
3
35
675
David Grenwood retweeted
Opus 5.5 leads all six new AA indexes, unless you count the fake Gemini
24
7
3
284
24,640
David Grenwood retweeted
A Columbia scientist told ChatGPT she picks coconut ice cream 70% of the time. It chose coconut for her 100 times out of 100. That's Sandra Matz in this TEDx talk. Her biggest worry about AI sounds small at first: that it will make us boring. Her logic: • Every choice is a trade-off between the safe favorite and something new • Recommendation systems and chatbots get rewarded for safe bets that keep you engaged • So they keep picking the favorite, and you slowly stop exploring Another test of hers: asked 100 times as a new customer, ChatGPT named one of Baskin-Robbins' two most popular flavors 96 times. Newer research adds a twist. In a 2026 experiment, people writing with AI produced more varied stories when they were paid for originality. The sameness follows whatever gets rewarded. Her fix: a dial on your apps, from "spot on" to "wildcard." 13 minutes, worth every one. Would you turn the dial?
9
8
29
694
David Grenwood retweeted
👇 Here’s your recap of our biggest AI updates from October 👇
257
248
55
3,730
996,911
David Grenwood retweeted
TOOK TWO DAYS OFF X. BUT THE PICKS KEPT COMING. A friend celebrated his birthday, and our group hadn't seen each other in a long time. So we finally got together, had some fun, and switched off for a bit. Not exactly how I like starting my week, but it is what it is. Meanwhile, you probably saw my weekend predictions. 8 picks in total. 7 shared publicly on my profile. 5 of those 7 hit. That's a 71.4% win rate on my published picks. Pretty happy with that result. I think my analysis was solid this weekend, even if not every prediction worked out. Tomorrow, the Champions League is back, with some really interesting matches coming up. Expect me to be much more active over the next few days. Time to get back to work.
21
2
81
1,222
David Grenwood retweeted
BREAKING: Anthropic has officially confirmed that its next AI model, Claude Fable 5.5 will be more powerful than its previous models. we are so back
27
13
4
623
68,693
David Grenwood retweeted
This is most likely the latest Fable 5.5 checkpoint. The video is pure code. Fable 5.5 will change the entire game. Im speechless. And yet, 99% of folks outside of our bubble have no idea. h/t @chetaslua nitter.gq/chetaslua/status/21171…
189
172
47
4,900
242,981
David Grenwood retweeted
i think we'll look back at claude opus 5.5 as the moment anyone who's ever wanted to become a vlogger, make internet videos, or build a video marketing machine could do it without a team or editing skills millions of people who would've sat on the sidelines will now be making videos over the next 24 months
274
71
32
2,272
218,506
David Grenwood retweeted
On Artificial Analysis, GPT-6.1 Sol uses about a third of Opus 5.5's output tokens at the same effort, and half if you account for score. In real T3 Code use, Sol spends 41% more output tokens per turn than Opus.
38
5
5
454
61,186
David Grenwood retweeted
Try Grok @Bot! You will love it. We land a few dozen fixes, features and sometimes powers to Bot 7 days a week and we look at your feedback to make Bot as perfect as possible.
本当にこれ マジでCursor(Grok Bot)を使っていない人は一度だけでいいから使ってみてほしい。 Claude Code、Codexよりも断然良いから。今UXの観点でCursorが負けているところは正直ないと思っている。 モデルの性能という点で現状はClaude Codeも使わざるを得ないが、Grokの性能がOpus 5.5と同等レベルになったらCursorオンリーにできると思っている。 SpaceXAIさん、次のGrokは頼みますよ。4.7も悪くないけどOpusと比べると力不足なので...
3,024
1,236
321
12,186
8,759,150
David Grenwood retweeted
Watch the window in the first 5 seconds. I made a 28-second Halloween film with Claude Opus 5.5. No video model, no editor. It’s one HTML file in three.js, sound included. Then I hid things in it. Some you won’t catch the first time. 🎃
15
4
35
777
David Grenwood retweeted
Someone built an entire game development studio inside Claude Code. 49 AI agents, 72 skills, and a full studio hierarchy working together to help you build games from scratch. Get started in 3 steps: * Download: Clone Claude Code Game Studios from GitHub * Launch: Open the project in Claude Code * Build: Describe your game and let specialized AI agents handle development You don’t need a 50-person team to start building your dream game anymore Repo: github.com/Donchitos/Claude-…
67
53
9
518
41,458
David Grenwood retweeted
openai naming their newest model after a dying star thats exhausted 99% of its energy and will soon implode
Our internal tracker picked this one up yesterday!
58
1,616
15
24,642
231,828
David Grenwood retweeted
Three weeks of training camp for Le Chonk. No days off. Thanks for putting it to the test, @ThibaultJaigu and @RequestyAI.
Impressed by the new @MistralAI large-4 update! The model update is much better at coding and performs as the 2nd best OSS model on our coding benchmark
114
66
17
980
51,405
David Grenwood retweeted
Introducing Gemini 4 Carbon, our latest frontier model. Carbon delivers improved performance across domains — ranking #1 on v5.0 of the @ArtificialAnlys Intelligence Index and sets a new state-of-the-art in cybersecurity and long-horizon capabilities. It’s rolling out starting this week to cyber defenders through our Fairwind Program, with expanded access coming soon 🧵
713
1,323
627
12,573
1,170,251
David Grenwood retweeted
No amount of iteration is going to turn a mediocre AI model into a frontier model. Everyone kept saying GPT-6.1 Terra was so cheap that it could outperform Opus 5.5 simply by iterating more. So that theory was put to the test. Opus 5.5 got ONE shot. Terra got the exact same API budget to iterate and improve its result. Same task. Same money. The result? Opus absolutely destroyed it. Terra is impressive at computer use and certain other tasks, but it’s NOWHERE near Opus 5.5 in coding. You can’t brute-force your way into intelligence.
Opus 5.5 is in a completely different league. I gave it and GPT-6.1 Terra the same one-line prompt: recreate Mickey Mouse in 3D. Terra was 17x faster, 45x cheaper, and used 9x fewer tokens. And Opus STILL absolutely destroyed it in quality. Sometimes you really do get what you pay for.
19
5
1
71
9,435
David Grenwood retweeted
GOOGLE IS SO BACK. GEMINI 4 CARBON JUST MOGGED OPUS AND ASTRA 💀 I added the two flagships to my Carbon eye test. Same prompt: Gemini 4 Carbon → 29m 49s Opus 5.5 → 43m 21s · $9.25 GPT-6 Astra → 22m 54s · $3.28 Gemini just made other models look old Waiting for Carbon on aimlapi.com
🤝 Paid partnership
GOOGLE'S NEW GEMINI 4 CARBON IS SCARY GOOD AT 3D 🥶 One prompt: a robotic eye that blinks, focuses, then explodes into every part. Gemini 4 Carbon → 29m 49s Waiting for it on aimlapi.com
41
19
5
344
32,376
David Grenwood retweeted
Opus is Opus.
30
34
6
976
126,937