Keubiko 🇨🇦🇺🇸🏳️‍🌈🏳️‍⚧️🇺🇦💪

Simple iteration on app development. This morning was just fixing bugs. After a few turns it just stops and says come back at 1pm. Extra annoying as it was fixing bugs it should have one-shotted.

I hear ya, but for $20 I get Gemini Pro which also gives me tons of storage, have rarely hit limits, models are way cheaper, and now they give me a $10 Google Cloud credit so my apps run free as well (for now). Just hard to pay 5x that for muh Claude. But it is better at coding no doubt.I’m cheap

15% tariffs on everything fixes this.

Claude Code is basically useless at the Pro ($17/mo) level. 20 minutes on then a 3-4 hour wait to use it for another 20 minutes.

Cap is way way too low. borderline unusuable without the Max plan I guess.

I think there’s an EDGAR API to pull in that data in XML format. Haven’t tried though.

I find that true with Gemini as well, but can solve it by telling it to calm the fuck down and take things one step at a time. My typical workflow now is to spend far more time on up front planning and specification, dump it into a .md file and use that to keep things on track.

Only after he proves that snorting cocaine off a toilet seat in a honky tonk at 3am is safe.

She should have said the Dow was at 48,000 and told him to get fucked, #BondiStyle

This was Claude Code and just some basic app prototyping. Used Opus (high end model) but JFC it feels like I can do 10x more with Gemini before it starts chirping me.

Just discovered this. Didn't even do much TBH. I almost never hit limits in Gemini. Booo.

So does the Twitter one though.

Excellent. Thanks. Brave new world out there.

Grok and to some extent Gemini have been trained on so much baggie content from Reddit and Xitter that the models think there was a massive short position etc - 99% of the MMTLP content out there is from drooling idiots so models can't seem to differentiate until challenged. Claude better.

Haven't played around with those yet. I know Antigravity is putting them in as well. I'm a coding charlatan so still getting my sea legs, but from what I've read they are great for quality as well as token usage.

When you come off a bender and start reading your 5-day old email alerts in order.

9/9 these are just initial impressions. Haven't even tried CoWork yet. My initial sense is that I'm likely to keep Claude, and am looking forward to having Sonnet and Opus in Antigravity (Google's IDE) which is what I make apps with for now. $17/mo bargain. Only scratching surface so far.

8/n not surprisingly, Claude Code appears very excellent. Seems quite a bit slower but that's less important than being more accurate in vibe coding. I'm very impressed. Like-for-like frontend prototyping seems superior also.

7/n The Claude Excel plugin is very cool, but was disappointed in its ability to build a complex model. My initial take is that it's very useful for simpler spreadsheet tasks. I haven't tried the Powerpoint add-in yet.

6/n 1M context window for Gemini is 5x Claude - have found that has mattered in some cases when uploading a bunch of docs for analysis.

5/n Gemini GEMS (custom persistent "personas" with info) have no apparent Claude equivalent.

4/n Claude has a "Projects folder" option which is great and it's bizarre that Gemini doesn't have it (yet).

3/n No image or video gen in Claude. I need that so at a minimum I'd be looking at keeping both.

2/n I find general Q&A with Gemini to be superior - better answers, Deep Research option, and Gemini Flash way smarter than Claude's fast "Haiku" which I find unimpressive.

1/n As a long-time Gemini user, about 5 days into an additional Claude subscription, here are my quick initial takes:

Dumbest baggies around, outside of $BBBYQ stupes.

Yes this is indeed what the baggies believe.

Claude gets it. @shorty-ocelot.bsky.social @tiggersdad.bsky.social @sunnycorax.bsky.social

Elon has now completely given up on pretending Grok is actually a competitive model on benchmarks or task performance and the entire release PR is just hyping how anti-woke it is