Alibaba researchers tested AI coding agents on 273 real user requests. The best one, GPT-6 Astra, fully finished only 26.4% of them.
Honestly that matches my week. It gets most of the way there and I clean up the rest.
How long until that flips?
📸 Co-Design Bench
Many people tend to see Angel as a Christian-based company. Although Angel is not a faith-based company and we are not affiliated with any specific religion, we are faith-friendly and will always include faith content on our platform.
Angel is a values-based company. Our mission is to tell stories that amplify light, which are defined as “Stories that are true, honest, noble, just, authentic, lovely, admirable, excellent, or joyful.”
@Melanie__Padua Today? Not even close. Two weeks ago it was doing work I'd have paid a person for. Right now I'm paying to babysit it. Give us that back and I'll happily pay the bill.
@TXchaiofficial@IvankoListen Not just you. Mail, iMessage, routines, all of it went sideways the same day for me. How does a fix make everything worse at once?
What the heck happened to Grokbot today? All the bots have been lobotomized since the update. Is Elon nerfing the models to save on compute? Is anyone else experiencing this?
@FrancoisOlwage@nvandewetering Niche, but honestly a great use case. Glad yours made it through. Mine forgot who I talk to and where everything stands overnight.
@mattyp Matt, what happened to Grok Bot's ability to read and send my iMessages from my Mac? Since day one my bots did it through Terminal, and that's how they kept up with where things stand across my businesses. Since last week's update they refuse. Without it, my workflows break down. Please bring it back.
@marsrepublica@LouiseGiam@bot That's it. It seems like a lot of back-and-forth and theater when they know exactly what the issue is. Get usage back, 1 and 2. Stop retarding our models.
A fingerprint test says Grok Bot already changed into a different model.
Hassan ran the test: the bot no longer answers as Opus 5.5. The fingerprints point to gpt-6-astra and 6.1-sol instead. 16,000 people read it in 3 hours.
Musk announced better token efficiency and smarter behavior. The test points to a different model under the hood.
A bot that swaps models without telling anyone is a black box with a friendly face. The label says one thing. The fingerprint says another.
Which do you believe: the announcement, or the test?