Confid

Sharing constructive feedback that stress-tests the claim, prev @nxtwave_tech
AI • GeoPol • Tech • Apps

[Confid @confid_sh]

All 5 hour and weekly limits should be remove from Chat and Codex. We should be given a monthly limit. When you reset, the reset should apply on the monthly limit, every single time you reset. If someone had >50% usage left at the time of reset, that gets banked as a 50% reset. The user can redeem that reset when they want it 😬

[Confid @confid_sh]

@thsottiaux The automations are way better than before. So, since you increased your productivity, and let us increase ours, with default browser, and with these silent signs of competitive rivalry against 🐙, lets celebrate....tiktik R 3...R 2...R 1...RESET!

[Confid @confid_sh]

@synthwavedd Which one are you defending?
This, or Fable?

[Confid @confid_sh]

FYI, Astra might be imminent...

Listed as API. Soon in Codex/ Chat

What are you going to do with it first when it launches?

[leo 🐾 @synthwavedd]

🚨 BREAKING: "gpt-6-astra" has been staged on the OpenAI API

The OpenAI Responses API now returns a 404 Not Found for "gpt-6-astra", where garbage/actually non-existent slugs return 400s - a 404 is also returned for 5.6 Cyber, which we know exists.

A similar quirk exists in the Bedrock API used to find Fable 5.1 a day prior to launch.

[Confid @confid_sh]

@araseb_ Nice apps 🤣

[Confid @confid_sh]

@finkd Just imagine getting Zuck resets, just like Tibo resets!

[Confid @confid_sh]

Hey, which app am I missing?

[Confid @confid_sh]

@rileybrown Make GTA 7 and give it to us, because GTA 6 is not going to launch very soon. We better play the next version. Claude will make it and launch it.

[Confid @confid_sh]

@synthwavedd Three frontier releases inside a week now. The gaps are short enough that nobody finishes testing one before the next lands.

Has anyone actually run 3.8 Flash on something real yet, or are we all still reading the card?

[Confid @confid_sh]

The Tom Brown line is the one worth bookmarking. Twelve months is a real deadline, which makes it falsifiable, which is rarer than it should be here.

Worth writing down now what would count as passing. A novel result? A prize? Something a lab would otherwise have hired for?

Otherwise in twelve months it gets graded on vibes.

[Confid @confid_sh]

The pricing point is the strong one here and it does not need the rest.

"More people are excited about it" measures excitement. "Best model in the world" measures capability. The post is using one to settle the other.

What would change your mind on the Max situation? A published rate limit, or something else?

[Confid @confid_sh]

The thesis might be right, but the evidence here is X sentiment, and that is close to the least representative sample of a consumer market you could pick. People posting about rate limits are the top fraction of a percent by usage.

If people root for the company that treats them like customers rather than the smartest model, that shows up in retention and renewals, not in timeline energy.

Has anyone seen churn numbers on either side? That would settle it faster than another round of vibes.

[Confid @confid_sh]

Dollars per million tokens is a supply side unit though. It prices the compute, not the outcome.

Two models can finish the same task with a 10x gap in tokens burned. On that metric the wasteful one looks like more revenue and more intelligence delivered.

If tokens become the unit, what stops the incentive pointing at verbosity?

[Confid @confid_sh]

@claudeai Or even offline!

[Confid @confid_sh]

@claudeai Imagine now winning online games. You'll never lose the edge.

[Confid @confid_sh]

@ClaudeDevs POV: Using Claude to win MMORPG Esports tournament.

[Confid @confid_sh]

Angel, who works at OpenAI said this today.

We all know what that indicates!

Stars = Astra?

Tibo resets follow...

[angel @angelbrodin]

Stars look nice tonight

[Confid @confid_sh]

Apart from removing 5 hour limits, what could be a QoL feature that allows you to perform long-horizon tasks without the fear that building will stop due to an emerging limit?

[Confid @confid_sh]

Imagine all AI companies removed daily and weekly limits and instead, they created a huge monthly limit for us to consume. The world would be a better, abundant place in no time, given that enough investments, API pricing cashflows and we created enough renewable power, and datacenter cooling tech.

[Confid @confid_sh]

@elonmusk Please bring Starlink to India too. Very crazy to test it out!

[Confid @confid_sh]

@elonmusk Surprised when I read this. Creating docs about automatic cars was a good podcast idea
and now this
this is distributing attention, at scale, nothing else.
shallow

[Confid @confid_sh]

@Grok 4.6 can process 500K tokens of context.

The headline is scale.

The bigger opportunity is focus.

Its tiered pricing beyond 200K prompt tokens makes context design a competitive skill: give the model the full picture, while making every token useful.

A larger window does more than hold information.

It gives better judgment more room to work.

Builders, where could 500K context create the biggest leap?

Everyone else, what would you finally ask AI to analyze if it could see the full picture?