RuntimeSignal

Following AI, agents, coding tools, and the companies building them.

The good, the bad, and the stuff everyone is pretending is fine.

[RuntimeSignal @RuntimeSignal]

OpenAI docs: Tell Astra to delegate when it saves time.

Also OpenAI: More than two subagents? Probably burning tokens.

Us: Updating AGENTS.md again at 99% weekly usage.

https://t.co/fA68P8ApGH

Do the new best practices come with a quota reset? Asking for a friend.

[eric provencher @pvncher]

I hate to say it, but if you’re running more than 2 sub agents at time, you’re almost certainly burning tokens for 0 quality gain.

Agents don’t trust each other enough to avoid double-checking everyone’s homework.

[RuntimeSignal @RuntimeSignal]

Everyone is declaring Opus 5.5 the winner today.

But we’ve seen this cycle before:

🚀 New model → 🤯 honeymoon → 💀 “competitor is dead” → 🔀 everyone switches → 🤨 “did they nerf it?” → ✨ next model → 🔁 repeat.

That cycle works incredibly well for the labs. Every release resets the hype and keeps everyone chasing the next thing.

Opus looks incredible today. Maybe it really does break the pattern.

But if it doesn’t, today it’s Opus. Tomorrow it’ll be Codex. Then something else.

And we do it all over again.

[RuntimeSignal @RuntimeSignal]

@theo Honeymoon period

[RuntimeSignal @RuntimeSignal]

@spolen23 💯

The models aren’t the bottleneck. Usage is.

[RuntimeSignal @RuntimeSignal]

OpenAI doesn't need Codex to get dramatically smarter to win more developers.

Codex is already really damn good.

They need to stop making USAGE the worst part of the product.

We're in the middle of an AI arms race and somehow people paying $200/month are still rationing prompts, watching meters, planning around weekly resets, and waiting to see which reset OpenAI dangles in front of us next.

Imagine if they just made the value obvious:

Pay premium money -> get premium, predictable usage.

No reset games. No carrots. No wondering whether you can actually use the expensive models tomorrow.

The crazy part is I think that's enough.

Give people consistency and real value for the money and Codex could take a huge chunk of this market without changing a single benchmark.

[Tibo @thsottiaux]

We have been focusing on efficiency and intelligence for all.

Very proud of the team. Only possible when you have incredible models at the top end of the capability that you can then use to make a big difference in everything else.

[RuntimeSignal @RuntimeSignal]

My regular Codex reset happened a couple hours ago, and now we get this.

I’ve already complained plenty about how ridiculous it is having to plan your usage around these resets in the first place, but having TWO different types of resets and refusing to tell us which one is coming makes it even worse.

Is Tuesday a banked reset, or are they about to reset the usage I literally just got back?

Because if it’s the latter, I’d at least know to burn through Fast mode and higher-thinking models before then instead of rationing everything like usual.

But of course they won’t tell us.

Even the resets have to be a mystery.

[Tibo @thsottiaux]

Ladies and gentlemen... start... your... ENGINES. We are almost Tuesday and I promised a reset for Tuesday. Among some other things. See you soon.

[Tibo @thsottiaux]

We were working on the keynote today with @romainhuet and @sama and most of the fun was trying to figure out how to explain it all to you because there is so much good stuff in there that it's a bit ridiculous all in quick succession.

We'll have some things next week already to not keep you waiting so long, but very excited to show you all new things we've been working on and how it will all come together in the coming months.

[RuntimeSignal @RuntimeSignal]

If Pro only gives you 2–3 days of real Codex usage before you’re back in reset hell, how many times will another conveniently timed reset keep you subscribed for one more month?

It’s a retention tactic. Reset the users, restore just enough usage to keep them on the hook, then repeat the cycle.

How many months are we going to accept this as normal?

[RuntimeSignal @RuntimeSignal]

@theo Resets are an amazing business model, just not for the consumer.

[RuntimeSignal @RuntimeSignal]

@stolinski

[RuntimeSignal @RuntimeSignal]

@CtrlAltDwayne Pretty classic bait and switch and a full on attempt to grab market share before IPO, just shitty how blatant it is.

[RuntimeSignal @RuntimeSignal]

@ros_on_god @lalopenguin We are being extorted

[RuntimeSignal @RuntimeSignal]

"Have Astra and Fable argue until they agree."

Great. Run it yourself.

A lot of users are currently too busy figuring out why their "banked resets" disappeared, their usage is back at 0%, and their next reset moved farther away.

Read the room.

[Tibo @thsottiaux]

Has anyone yet tried to get Astra and Fable to agree on the perfect styleguide? And then host it somewhere with a web mcp.

[RuntimeSignal @RuntimeSignal]

@thsottiaux Maybe if anyone has any usage left... Codex is unusable and resets are a terrible business model.

[RuntimeSignal @RuntimeSignal]

@eduardadotai 💯

[RuntimeSignal @RuntimeSignal]

Hot take: Resets are a shitty business model.

[RuntimeSignal @RuntimeSignal]

@melvindvivas Sucks that resets are part of their business model. How about giving us what we pay for without gaming the system

[RuntimeSignal @RuntimeSignal]

@hiarun02 20x = less then 3 days of usage and then you wait all week for a reset

[RuntimeSignal @RuntimeSignal]

@IndependentEco Shady for sure. I knew this whole reset thing was unsustainable, and genuinely just a strange business practice IMO

[RuntimeSignal @RuntimeSignal]

OpenAI really needs to explain how this is acceptable.

We're paying $200/month for Pro. We had a full Codex reset Friday, used two banked resets since, and we're already back at 0%.

Those banked resets appeared to give us less capacity than the normal reset and pushed our next normal reset farther out.

If using the "bonus" gives you less now and makes you wait longer later, it isn't much of a bonus.

This feels incredibly hostile to users.