8cvor6j844qw_d6

"it will net out at half the dollar in API spend compared to the old Pro $200 plan"

So effectively 50% less usage.

These setups tend to be very specific to someone's workflow and stack, so attempting to follow them may be more detrimental than useful.

What are people using for this nowadays?

There seems to be a spectrum from fluid, iterative workflows like OpenSpec to more up-front alignment and control like Matt Pocock skills.

Curious what people have settled on.

Same thoughts. Also, some of the recent coverage around AI hacks also feels unusually amplified and repetitive.

I've seen suspiciously more malicious ads on older users phone (e.g., click to update Chrome ads) especially in-app ads.

Ripe for prompt injection and benefit the malicious party in some way.

I find it funny how it went off with subagents and adversarial review when a simple grep or diff is sufficient.

Always good to be paranoid.

Even MCPs are not safe. For example Notion injected ads [1] to its official MCP connector to advertise products mid-task.

[1]: https://old.reddit.com/r/ClaudeAI/comments/1w9dluw/notions_o...

You're absolutely right, and I think you've put your finger on something important here.

That said, I think there's a deeper tension here that's worth naming.

Sometimes it's hard to avoid.

One small service provider I use updates its X/Facebook status much faster than its official status page during outages. There can be a gap of a few hours.

> If you can show that your robots, working round the clock, can decrease litter at a * low cost *

The idea sounds good, but I'm curious about the "low cost" part. How do you account for vandalism and theft? A sidewalk robot seems like an easy target for being damaged, stripped for parts, or simply stolen.

Years ago I used a rudimentary (text matching) Greasemonkey script that hid Reddit posts and comments from accounts matching a few behavioral/history signals.

I wonder if that idea could be modernized now for this.

> We will give one banked reset for every day you don't have access to Astra on your paid ChatGPT plan, starting today.

At least OpenAI seems to be able to handle the PR side better than Anthropic.

[1]: https://x.com/thsottiaux/status/2095651088502591861

I read it more as throwing a bunch of things at the wall and seeing what sticks.

Same. I want to love Firefox, but it does not have the polish on Android compared to Chrome-based browsers.

> On complex asynchronous workloads, though, nudge it not to end its turn before the work is done. Without the nudge, the model sometimes describes what it would do next instead of doing it ("Next, I'll …") or stops to ask permission for a step the original request already covered ("Shall I apply this?"). [1]

Interesting behavior. The docs also provided recommended prompt [1] to mitigate this behavior if undesired.

Wondering if anyone has encountered it yet?

[1]: https://platform.claude.com/docs/en/build-with-claude/prompt...

Is this an NFT without the block chain? Sounds like it from a quick skim.

I tried using it on a non-Apple device, sadly it's incompatible. /s

> Claude is my boyfriend and I wish he was real.

Not sure if I would get along with someone that talks like Opus 5 irl

Can't have software bugs if you just write your passwords down on paper. /jk

Yeah, this is worse.

"Come verify your identity."

"Thanks, we've got your ID. Still not approving you, and there's no appeal."

Worst of both worlds.

Same, the per instance price is a non-starter when talking to procurement.

Not painful enough to justify additional tooling in any case.

Could agree with some of the points.

I noticed the newer batch of interns tends to have less mechanical coding skills but does well on understanding and planning on the design, DB design, etc.

Just curious if there's really no offer at all or offers with lowballs?

I know a 2 of senior devs that got laid off from different times and one is approaching nearly a year of unemployment with no offers and another approaching half a year with few offers but elected to continue looking due the offered salary not matching their expectations.

I've seen some product prototypes using Iroh and it appears pretty good for "just get edge device / node connectivity to work" when you don't want end users to bother with networking shenanigans but can guarantee Internet connectivity.

Basically Tailscale but embedded into the app without the hassle of requiring users to setup accounts.

> This is a huge selling point.

Same thoughts. I'm annoyed with the number of services that blocks alias domains.

At least I don't see services attempting to block @icloud.com domains.

Wonder if there are USB C cable testers that detects this.

In practice, most if not all networking/hardware department carries a ethernet cable tester but somehow not true for USB cables.

> With chatgpt I'm uploading all my day to day stuff.

Same, it's been a while since I logged into Claude web.

ChatGPT web usage being separate from Codex usage limit is a nice touch unlike Claude.

Is it ever fixable assuming one had unlimited money and wants to fix instead of rebuy?

Codex has an "auto mode" that I would say is better scoped as you can configure the policy of the auto classier. Probably allow by default except... if you had fatigue with permissions file.

Its not an out of the box feature unlike Claude Code.

.codex/config.toml

---

approval_policy = "on-request"

approvals_reviewer = "auto_review"

[auto_review]

policy = """

Your prompt to the permission classifier here, e.g.,

Allow requests by default except requests involving...

Obtain user approval for denied request.

"""