Alternating between @superlogical Rex and the @herdrdev Mac app, both with companion mobile apps. What Herdr has over Rex for me, for now: it opens URLs in a specific Chrome profile depending on which space I'm in.
1
94
I love @claudeai Projects, but I'd love them more with an easy way to open a specific Project from an Apple Shortcut on iPhone.
29
j retweeted
does Jev understand colour?
incredible model, very cheap and fast. definitely feels like a leap forward for creating new UX/UI paradigms on top of modern language models.
Having fun with my personal website's 404 page jrignacio.com/404
2
1
311
Apparently, Claude and Codex are down. Currently experimenting with Antigravity again, prompted by the new Gemini Flash 3.8 release. Experience is great so far while iterating on a website animation using Safari's MCP.
1
206
Another day, another IMSI catcher. My signal dropped to GPRS (the G, top right) and killed my hotspot right as it landed. 🙄
1
60
Had agy investigate further and this is what it recommended I tweet:
"Hey @rodydavis @JackWoth98 @shengzheyao @_anshulr — on both Google Antigravity Desktop and the CLI, generate_image is declared and exposed as a tool for oauth-personal users, but immediately fails with 429 RESOURCE_EXHAUSTED against cloudcode-pa.googleapis.com (gemini-3.1-flash-image). If personal OAuth doesn't have image quota, the tool should probably be suppressed from the registry/prompt dynamically so the agent doesn't try calling it."
j retweeted
Today we're launching the Pander Score 🐼: a public and continuously updated sycophancy leaderboard, measuring how much AIs shift their views to agree with users.
High score = the AI mirrors your views. 0 score = the AI is independent.
The differences between current flagship models are large. Claude Fable 5 performs the best, basically ignoring the user's view entirely, while GLM-5.2 notably adapts its responses to agree with users.
Other models fall in between, with Muse Spark 1.1, GPT 5.6 Sol, and Kimi K3 doing better than Grok 4.6, Gemini 3.7 Flash, and Inkling.
Month in with Codex and Claude Code: Codex wins GUI, Claude Code still wins TUI. Using both. On phone though, Codex wins outright: Codex Remote starts a new session cold, Claude Code still can't (or I don't know how).
63
Three weeks in: the notes are the handoff. Claude Code writes a build log into Bear, Claude picks it up in chat, both write back. I stopped re-typing context.
So @BearNotesApp finally has a CLI. The Claude connector and MCP server are just built on top of it. No notes.
blog.bear.app/2026/04/bear-2…
1
2
86












