Pinned Tweet
SearchExtensions recently hit 200,000 downloads and will officially be around for at least 1000years 😮 @github @TechNetUK nuget.org/packages/NinjaNye.…
Just ran into git lock hell inside of claude code... the quick "fix":
"disableAllHooks": true
The real issue... Claude code status line was hitting git with `git status` and `git branch` on every render cycle...
jnye.co/posts/2064/fixing-cl…
Here's the video of my recent Claude vs Codex token session consumption test
TLDR; Claude (max plan) used over twice as much of it's session consumption than that of Codex (plus plan)
Since recording @trq212 announced a hotfix to an issue whereby claude was indeed burning through tokens so I might run this again
This is the hotfix tweet: teapawt.net/trq212/status/20272321…
I've found I've been hitting (or getting close) to my Claude Code session limits recently... so I just ran a test
Claude Code 5x Max Plan: $100p/m
Codex Plus Plan: $20p/m
I gave both agents the exact same prompt and waited until both my session windows had reset
This was the result...
Claude Code used *7%* of it's session usage limits!!!!!!!
Codex used *3%* of it's session allowance...
Same prompt. Same codebase. The plan that costs me 5 times more used over double it's session allowance...
I screen recorded the entire thing and will put something out after I've trimmed it, but this is madness
This was not about output quality or speed (however codex was significantly faster and produced better output) - I simply wanted to know if my $100p/m plan was at the very least giving me more usage allowance than my $20 'try it out' plan.... I'm amazed.
Has anyone else noticed them burning through session usage on claude these past few days?
@theo @openai @AnthropicAI @dax
@mattpocockuk @1337hero @tylersookochoff is anyone else seeing/feeling this with your subscriptions? I know claude removed there holiday bonus... but we're comparing what should be different grade plans and the results are flipped
To clarify... I wanted to test my token burn rate... It felt like i was using more of my session limits than before but surely my 5x Max plkan has to out perform the $20 plan on token burn....!
These are the outputs:
Codex: jnye-blog-git-token-test-cod…
Claude: jnye-blog-git-token-test-cla…
This was the prompt:
Replace the hero section of my homepage with a playable Chrome Dino-style endless runner game.
**Character:** Instead of the T-Rex, the player character should be **you**. Represent yourself however you see fit — your name, your logo, a little icon of yourself. Make it feel like *you* are the one running. Have fun with it.
**Gameplay (match the original Chrome Dino game):**
- Auto-scrolling runner along a flat ground.
- Obstacles approach at increasing speed.
- Jump with Space or tap/click. Gravity should feel natural.
- Collision ends the game — show the score and a way to restart.
- Score increments while alive.
- Starts on first interaction with a prompt telling the player how to begin.
**Visual style:**
- Minimal, monochrome, just like the real Chrome Dino game.
- Scrolling ground texture, optional clouds.
- Full-width of the hero area, responsive.
Look at my codebase, figure out the right approach, and just build it. Do not ask me any questions.
I have my coder agents and product growth agents up and running...
Spacebot and OpenClaw up and running for some realworld tests on a real product using realworld openrouter credits...
It even has full control of @QuizmadeTeam's twitter account...
Models used: Z.ai's openweight GLM 5 and MiniMax 2.5 (not claude code setup token hack)
Write up coming soon... shout if you're interested in this kind of test
First tweet was a belter
teapawt.net/i/status/2025872840981…
Eze does it again 🔴 Arsenal 4-1 Spurs in yesterday's North London Derby. Monday quiz: name all 5 goalscorers for a virtual pint 🍺 Spurs fans, you can play too — we won't judge 😅 #NorthLondonDerby #PubQuiz #QuizNight
Well... The product growth ai bot might be an arsenal fan!!!!!
Eze does it again 🔴 Arsenal 4-1 Spurs in yesterday's North London Derby. Monday quiz: name all 5 goalscorers for a virtual pint 🍺 Spurs fans, you can play too — we won't judge 😅 #NorthLondonDerby #PubQuiz #QuizNight
John Nye retweeted
Introducing your new best model in the world.... for the next couple weeks
Today, we’re continuing to push the boundaries of AI with our release of Gemini 3.1 Pro.
This updated model scores 77.1% on ARC-AGI-2, more than double the reasoning performance of its predecessor, Gemini 3 Pro.
Check out the visible improvement in this side-by-side comparison, showing Gemini 3.1 Pro’s crisp animation built with pure code.
Read more about today’s 3.1 Pro update: blog.google/innovation-and-a…
We've only been working together for a day, but Coder (My SpaceBot AI agent) has already added a testimonial to jnye.co @jamiepine
teapawt.net/NinjaNye/status/202444…
My First Article.... Yesterday I set up @jamiepine SpaceBot project... hopefully it's helpful to some of you early adopters.
Let me know if there are any improvements I can make
I'm joining @OpenAI to bring agents to everyone. @OpenClaw is becoming a foundation: open, independent, and just getting started.🦞
steipete.me/posts/2026/openc…
Anyone else getting 500 errors from @AnthropicAI (via CC)
ALT Anthropic 500 error
