Last Friday I was curious to give Codex another try after 6+ months of using Claude Code exclusively.
Was not expecting this:
Codex absolutely blows Claude out of the water. I havenโt even opened Claude since.
Codex has really made me realize what a sycophant Claude is.
Over the past month, some of you reported Claude Code's quality had slipped. We investigated, and published a post-mortem on the three issues we found.
All are fixed in v2.1.116+ and weโve reset usage limits for all subscribers.
@manuel_frigerio ๐คฃ I wasn't trying to convince you. I appreciate that link and hope it helps! Mostly just wanted to share some funny (in hindsight) examples!
Anyone else underwhelmed by Opus 4.7?
It's slow and seems to make more incorrect assumptions and silly mistakes than 4.6 did.
The biggest change I've noticed is that it's more confident and persuasive when making mistakes.
@manuel_frigerio Another example: even after explicitly telling it multiple times to use our /prod-logs skill (which uses `scalyr`) it insisted the command didn't exist, then admitted it exists but started making syntax errors.
Then it admitted it basically guessed rather than reading the skill.
@manuel_frigerio Fair, I'm sure adjusting my prompts could help somewhat...
...but, for example, this is what it did when I asked it to write me a Ruby script ๐คฆ๐ปโโ๏ธ
Boss: Lunch is on me. Any ideas?
โPizzaโ
โSushiโ
Me: (suspicious that John in HR is secretly an implicit ActiveRecord callback firing somewhere deep in the model layer)
โafter_commit?โ
Me: (staring at John to see his reaction)
@GregorySchier Not crazy! I briefly considered this before deciding to abandon Linux (for now).
Is it possible to get third party apps (Firefox/Chrome, Slack, etc) to look & feel native? Thatโs the part that truly irks me.
My sister has pancreatic cancer. Sheโs fighting like hell but itโsโฆ a lot.
Sheโs used up all her sick days with several days of treatments and procedures ahead.
The school she teaches at launched a fundraiser to help and I want to share it for reach.
https://t.co/udWQIeTLnf
@adamwathan@jorgemanru@OmarchyLinux I tried (again) a few months ago and Omarchy is awesome...
...but I just can't untether myself from Apple Messages and Photos, or the seamlessness between my MacBook Pro and my iPhone ๐ข
I've been wishing for this exact same thing: skillsets.
I have some Rails projects that use Hotwire + Minitests and others that use Inertia and RSpec. I don't know of a good way to enable/disable conflicting skills without resorting to project-scope skills.
Scripts + symlinks?
We might need a virtualenv for Claude Code for switching installed sets of plugins. Some overlap and it gets confusing. Is this something on your radar @AnthropicAI ?
@bradgessler I chose the first option, but I typically don't add `account_id` everywhere unless denormalizing for performance.
I scope all queries off the tenant (Current.account.projects) to the point that querying models directly (Project.where) becomes an anti-pattern in the codebase.
โจ A dream I had finally came true: I can now chat directly with my sites to build any feature or fix any bug just via Telegram
I've been playing with OpenClaw for 3 weeks now and it's great but I was always too scared to run it on any production server
And I was right a bit as @marckohlbrugge was able to hack it by social engineering and acting as if it was me, and with enough tries it believed him, and was able to modify the server, change SSH keys etc. of course I had it isolated properly on its own VPS and it didn't touch anything sensitive (as it should!)
Marc then reported that bug to @steipete who patched it fast
But I wanted to try something more basic and simple, and I think maybe more secure: to just connect Claude Code on my server to Telegram which would be hard locked to only messages from me
So I installed claude-code-telegram by @RichardAtCT on the server and run it as a system daemon and it works really well
The cool thing is that I was already using Telegram for server errors like this:
> Photo AI - โ Random credits giveaway failed (Attempt 30/30) with an exception: SQLSTATE[HY000]: General error: 5 database is locked
So now I can just reply, "Ok fix this", and Claude Code on the server in production will try (and probably succeed) in fixing it
In the video below I asked it to make show [๐ณ Parks ] on the map by default on load, it did that, then I reloaded the page and it instantly worked
One thing it still needs is sending actual messages while it's doing stuff which OpenClaw does really well, it's annoying to just wait while it says "Working..." but that's probably next
@erick_tmr@claudeai Yes, I understand the code is probabilistically generated. And when it suddenly starts consistently generating dramatically terrible results over the course of weeks when no other variable has significantly changed, Iโm inclined to suspect the black box is the problem.