Pinned Tweet
I actually prefer desktop Claude Code and Codex apps
But Juggling all the accounts and sessions?
- Not so much!
So I built AI Juggler:
an open-source management and orchestration tool that works WITH the native apps, not instead of them.
Obviously that wasn't enough and I got carried away and gave my agents a virtual 3D office.
If we think of AI as sentient, might as well give me them an office, right?
Have you run into this problem, how many agents are you juggling at once?
I made the last 2 years of AI coding benchmarks into a Tron style race, week by week!
Care to guess who held the crown the longest?
...and who had the longest winning streak?
Was it Claude, ChatGPT, Gemini…
or maybe GLM or Muse?
Guess first, then watch.
34
Thanks to AI Coding Agents, now I have these two options:
A: Play GTA VI + Claude Code building my product
B: Post on X + Claude Code building my product
Which?
3
6
336
I've used Astra 6.2 and Sol 6.2 on a brand-new real-use project I was building:
- Product planning & research: Poor
- Architecture: Average
- Coding: Average
- QA / Review: Above Average
Honestly, at this point, is the only way for GPT Astra to catch up to Claude…
...by distilling Claude?
I was very surprised how poor its product development, research and planning was.
Especially presentation of the plan in comparison to opus.
Next to check Google Neon?
5
7
2,255
In AI coding, every % matters.
"Only 3% better" can be the difference between a model looping for days…
…and solving the problem in 20 minutes
I've seen it firsthand, multiple times
Ever had a "slightly worse" model just stuck there, promising but never delivering?
1
1
64
AI Juggler's 30-day usage: 4.5B new tokens, 198B more re-read from cache, 7 accounts.
At API prices, that's ~$63.9K.
My workhorse?
Claude Opus 5.5, 3.7B of those tokens.
Sorry, but I just can't trust GPT Sol or Astra with anything serious yet.
What's your workhorse model, and what does your token usage look like?
1
65
If only AI Coding Models could get these kind of scores!
GPT, Claude, Gemini ???
GTA 6 review score roundup:
• IGN - 10/10
• VGC - 10/10
• Empire - 5/5
• Forbes - 10/10
• Eurogamer - 5/5
• ComicBook - 5/5
• Digital Spy - 10/10
• TechRadar - 10/10
• GameSpot - 10/10
• ScreenRant - 10/10
• Destructoid - 10/10
• PushSquare - 10/10
• CGMagazine - 10/10
• GamesRadar - 10/10
• DualShockers - 10/10
• Game Informer - 10/10
• EDGE Magazine - 10/10
(Metacritic: 99)
84
Hello Book Authors,
I'm looking for writers to beta test a professional manuscript editor and a world building tool. It is a very advanced writing tool.
In exchange, you will get lifetime full business access to the tool and our book publishing platform, BookAuth (sell directly, website, newsletters, ARC, writing tool)
If interested, please reply to this post.
#WritingCommunity #writerslife #AuthorsOfTwitter
3
3
145
Here are some links to check out the features:
bookauth.com/features/manusc…
bookauth.com/features/manusc…
20
People started comparing Claude and GPT with 3D rendering benchmarks
And voila, guess what:
OpenAI's next models, Astra 6.2 and Sol 6.2, show up with better 3D scores
Look, I have nothing against better 3D rendering
But it's really a tiny fraction of what we actually use these models for
Does this remind anyone else of the old hard drive and camera megapixel wars?
1
1
149
I keep seeing devs and founders lash out at their users... ...especially their beta users.
Calling them stupid. Mocking their complaints.
I can't believe it!
What the fuck?
A frustrated user is feedback.
Yes, complaining IS feedback.
That's your golden chance to make your product better and win them over.
Instead, you're letting your emotions win!
So GPT-6.2 Astra and GPT-6.2 SOL just dropped.
If you are running AI Juggler...
... you can just easily switch your Desktop or CLI session from Claude Code or other tool and continue your work in Codex to try out new 6.2 models
All this with only few clicks!!
Some of you have been asking very politely. Some of you have just been asking.
GPT-6.2 Astra and Sol are now available.
Super efficient and built to perform where it counts—the real world.
openai.com/index/introducing…
1
116
Check out AI juggler , an open source project:
figurehead.ai/ai-juggler
12
I asked Claude Code to write me a function for success.
This is what it came back with:
You guys do still remember how to read code. Right?
... and how do we debug the luck part?
3
1
146
Trust but verify... applies to models too.
Microsoft researchers showed that a AI coding model can be poisoned to behave normally...
until a hidden planted trigger makes it run a malicious command.
This is yet another reason why I don't trust the first few passes from any AI coding agent
Rule:
Write with one model
Review with another
3
2
101
Microsoft research article if you are interested in the details:
microsoft.com/en-us/security…
47
Which Black Friday AI deal would you rather see drop next week?
A: Claude Opus 5.6 + 50% off Claude Max/Ultra for 6 months
B: GPT Astra 6.2 + 50% off ChatGPT Pro for 6 months
Pick one, and why?
4
4
464









