UK-based AI commentary & culture. Big news, leaks, rumours, models, policy, AI tools, access & creative shifts. • Creative @Axylusion • News @koltregaskes2
UK
Joined March 2008
- Tweets126K
- Following7.2K
- Followers20.9K
- Likes209K
Pinned Tweet
A couple of years ago I predicted GPT-7 or 8 to be AGI by 2027 or 2028. It's looking quite likely so it will be a very big fluke if that's correct.
But with hallucinations, current accuracy and jagged knowledge still very jagged, we still have quite a way to go.
But for the labs 100% in-house is clearly the way. You need control of the full stack. In reality this is still probably only the OS providers so Anthropic and OpenAI need their own operating system.
OpenAI models coming to Gemini Agent and Antigravity soon, after being added to Google Cloud yesterday. This is for enterprise only for now.
If you use PHP please take a look at this urgently. Stay safe out there.
If you are wondering whether your website is affected, we have packaged a Codex skill to help check.
Point your Codex at this repo to guide you through it: github.com/openai/php2shell-…
Kol Tregaskes retweeted
Gemini 4 Carbon release is very soon, closed testing started and it's already in LM Arena.
Only the other day I said Anthropic were good at worldwide releases, then today they release a US only waitlist.
Why not make this invite only?
I think the slowdown would naturally come if we had a single point of contact like we have with the personal AI assistants that have recently arrived.
Claude Bot?
Kol Tregaskes retweeted
AethrBooks is now fully launched! 📚
A home for readers, authors, comic creators and the worlds they build.
Here’s what you can do:
📖 Read your way
• Books, comics and graphic novels
• Interactive 3D pages
• Custom reading pages
• Bookmarks, reading progress and milestones
• Samples of paid books before you buy
📚 Build your library
• Your own virtual 3D Grand library space
• Personal shelves and reading lists
• Dedicated comic and audiobook sections
• Purchased books are yours forever
• Discover books through the Daily Shelf, series pages and trailers
✍️ Publish your stories
• Upload books or comics in any format
• Interactive author studio with a live 3D book preview
• Custom covers, categories and tags
• Optional OCR for scanned manuscripts
• Free or paid listings
• Release dates and countdown pages
• Announce a book before uploading the manuscript
• Paid preorders, wishlists and release reminders
• Add a book trailer and edit your listings
• Control reviews, ratings and AI companion permissions
• Pen-name profiles and account switching
Don't like AI? Well we do but you can switch it off and customize your entire experience. Optional opt-out options for Authors.
🎁 Give someone their next favourite book
• Gift to an AethrBooks account, email address or private claim link
• Personal gift messages and a wrapped-book presentation
• Purchase share cards and buyer notifications
🤝 Meet your people
• Author profiles with shareable links
• Author directory and discovery cards
• Follow authors and receive notifications
• Profile conversations
• Signal posts with photos, polls and comments
• Photo and video Stories that disappear after 24 hours
• Ratings and reviews
• Hide posts or books and block accounts
🎬 Create beyond the page
• AethrCreate with an AI Director
• Book trailers from scripts, descriptions and reference images
• Simple controls, with advanced options when you want them
• Lip-sync options
• Image generation and a creation library
• Audiobook generation with narrator choices and the full credit price shown upfront
💬 Your reading companion
• Ask questions about the story and they will read along with you
• Customise its name, personality and profile picture
• Available for supported books and comics
• Respects reader AI preferences and the author’s permissions
Follow @AethrBooks
Your words stay yours. Your work stays private until you choose to publish.
Explore aethrbooks.com
No AethrPro till next week but as we are community based if you want to support AethrBooks and see it become the best, you can donate here: ko-fi.com/grimfel
This is the beginning and there’s more to come.
OpenAI, why do your sessions take so long to load in the desktop app? Long sessions sit there loading the model, and I can't interact with it until that's done. Then the conversations themselves sometimes load in chunks (both old chats and new chats) . It's also just very slow-moving. I'm talking local sessions here.
I'm not seeing this with Claude anymore. I'm not saying Claude's fast but it's definitely faster, and Grok is way faster.
I'm also still getting errors and having to restart the app quite a lot. It's starting to fail a lot more again. We did have a little period where it was quite stable, but it's going backwards again, I'm afraid, for example:
Kol Tregaskes retweeted
For less than 24 hours I fell in love with Copilot and I swapped to Opus 5.5 but today I had to turn it off because we weren't too sure about data protection.
And instantly my machine slows down. Output slows down. The errors come back and it's back to misery. Why are GPT models so bad?
Maybe it's a coincidence but right now it's pretty clear: swapping to GPT-6.1 Sol from Opus 5.5 has dramatically slowed down my machine. I don't get it because Copilot is in the cloud. 🤔
Gemini 4 Carbon release is very soon, closed testing started and it's already in LM Arena.
Because our regular chats are not locked...
Keep sensitive conversations out of view with Locked Chats in ChatGPT.
Move chats out of your regular history and search into a protected space you can unlock with a PIN on web and desktop, or a PIN, Face ID, Touch ID, or device unlock on supported mobile devices.
New messages sent in locked chats won't update the memory used in your regular chats.
Rolling out starting today to Free, Go, Plus, Pro, Business, Enterprise, and Edu plans.
I don't quite follow the Gemini 4 Argon weekly limit. It says 0% here, but when I select the model in Antigravity, it immediately says my quota will refresh on the 25th of October.
I had already waited a week for the reset on the 18th of October, when I didn't even get a chance to send a single message to Gemini 4. It just immediately came up with quota reached.
What on Earth is going on with Gemini 4 in Antigravity?
It's also confusing because I have AI Ultra in Gemini Web because I'm a tester, but AI Pro in Antigravity. I think it's getting a bit confused
For less than 24 hours I fell in love with Copilot and I swapped to Opus 5.5 but today I had to turn it off because we weren't too sure about data protection.
And instantly my machine slows down. Output slows down. The errors come back and it's back to misery. Why are GPT models so bad?
Maybe it's a coincidence but right now it's pretty clear: swapping to GPT-6.1 Sol from Opus 5.5 has dramatically slowed down my machine. I don't get it because Copilot is in the cloud. 🤔
Kol Tregaskes retweeted
OpenAI, how many types of sessions do you have now? I count four.
1. Work mode
2. Chat mode
3. Codex?
4. Dots
You were supposed to merge these into one not add an additional one. LOL.
And can all sessions please be in the cloud but you can give some sessions local access?
Kol Tregaskes retweeted
Is there a fundamental flaw with ChatGPT right now?
I ask my ChatGPT sessions to write todos at the end of every message and get those completed. In theory it should complete all those todos and have them ticked off each message but that's not the case. I'm finding it ticks off one, lists a whole bunch of others, which are unticked, which is fine if they eventually get done. But when I get the next message after this, those unticked todos just disappear and gone forever.
This is the experience I'm getting from running dots. I don't know if there's an issue within my workspace or this is a fundamental flaw with ChatGPT but tasks are not getting completed and they're actually just disappearing off the list.
OpenAI need to wake up fast. I'm about to dump it and move over to Grok bot so I can get some work done and not have to repeat myself a thousand times.
Is there a fundamental flaw with ChatGPT right now?
I ask my ChatGPT sessions to write todos at the end of every message and get those completed. In theory it should complete all those todos and have them ticked off each message but that's not the case. I'm finding it ticks off one, lists a whole bunch of others, which are unticked, which is fine if they eventually get done. But when I get the next message after this, those unticked todos just disappear and gone forever.
This is the experience I'm getting from running dots. I don't know if there's an issue within my workspace or this is a fundamental flaw with ChatGPT but tasks are not getting completed and they're actually just disappearing off the list.
OpenAI need to wake up fast. I'm about to dump it and move over to Grok bot so I can get some work done and not have to repeat myself a thousand times.
Kol Tregaskes retweeted
Fable 5.5
Gemini 4 Carbon
GPT-6.1 Terra
All coming this month.
And we have new GPT image models, new Nano Banana Pro language models, new Gemini Omni update and a new MiniMax Pro model all on the horizon. With GPT 'Bel' as a Christmas present. 😀
What did I miss?
We're getting Fable 5.5 and Gemini 4 Carbon this month according to The Information
But also GPT-6.1 Terra
So either Terra is now extremely good and capable, surpassing GPT-6.1 Sol, which wouldn't make sense, or it's actually over for OpenAI
I'm sure they know how much people want a new SOTA model from them and that a new Terra model worse than GPT-6.1 Sol would destroy all faith in them on X so let's wait and see what happens, they could also be releasing a new Astra at the same time