Eric Phy

Prod & Des @PickTheOdd | prev Prod & Des @StratzDota | Relearned to walk in '23/24

[Eric Phy @EricPhy]

@Porkbun @Tom_8496 @devdiary0x Oh whoops, I read @tom_8496’s msg as “how do _they_ make money…” I was referring to how registries make money, not you guys. You guys are the shit

[Eric Phy @EricPhy]

@Tom_8496 @Porkbun @devdiary0x Costs per domain are minuscule.
Revenue per domain are higher than minuscule.

[Eric Phy @EricPhy]

Construct 3 has been getting better every week since forever. Few platforms have ever achieved such an extraordinary release cadence over such a long duration. And here they are, with another massive banger!!

[Construct @ConstructTeam]

Launching Construct Game Services:
https://t.co/pHHnfHrIF0

A big new service from us we've been working on. Enhanced free plans with all Personal + Business #Construct3 subscriptions. Full Rest API also available.

[Eric Phy @EricPhy]

@ConstructTeam Huge!!!

[Construct @ConstructTeam]

Construct 3 and Construct Animate stable r487 are now out! Including:

⭐ A new 3D editor view
⭐ Lots of 3D model improvements
⭐ 3D bullets
⭐ Open additional Event Sheet Views
⭐ Pixel snapping
⭐ Single EXE exports
⭐ Lots more!

Full release notes 👇

[Sketch @sketch]

Would love to see you over at https://t.co/9CeRqe54I6 if you’ve got feedback on this one!

[Eric Phy @EricPhy]

@GaelBreton I found several bot accounts that sounded super legit playing contrarians in anti-4.7 threads. They even would pigback off each other. Ominous.

Aka: bot accts saying 4.7 is good and the naysayers weren’t reporting correctly, or were otherwise at fault.

[Eric Phy @EricPhy]

@TiffanyFong

[Eric Phy @EricPhy]

@Grummz 208/min is rookie numbers

[Sketch @sketch]

🆕 Things you’ve been asking us for 🆕

Selection colors, independent borders, corner smoothing controls, a new eyedropper with Color Variable support, and 150+ improvements and fixes.

Here’s what’s in our latest update — Dublin 🧵 ↓

[Eric Phy @EricPhy]

@3266miles @AnthropicAI This happens with every model, at least for gpt and claude, in the last ~2 years. Not sure why people are still gaslighting, and only novices should be finding this out for the first time.

But yeah, it sucks. It’s part of how these companies operate, for reasons unknown.

[Eric Phy @EricPhy]

@0xCharlota My concern regarding this is how modular and systemic it is. Just a bunch of “change this/that” cmds is going to result in horrible software very quickly.

[Eric Phy @EricPhy]

@sketch @emanuelsa @Apple Not sure crazy/misfit/rebels are the apple way anymore, but I’d love to see that energy come back. Doing my part

[Eric Phy @EricPhy]

@ConstructTeam Huge!

[Eric Phy @EricPhy]

Still begs the question why an independent researcher had to go to these lengths to discover bugs that are affecting a huge percentage of Claude users.

And this is about the 4th time this has happened in less than 2 months.

[Alex Volkov @altryne]

PSA: If you've been running out of Claude session quotas on Max tier, you're not alone. Read this.

Some insane Redditor reverse engineered the Claude binaries with MITM to find 2 bugs that could have caused cache-invalidation. Tokens that aren't cached are 10x-20x more expensive and are killing your quota.

If you're using your API keys with Claude this is even worse. This is also likely why this isn't uniform, while over 500 folks replied to me and said "me too", many (including me) didn't see this issue.

There are 2 issues that are compounded here (per Redditor, I haven't independently confirmed this) :

1s bug he found is a string replacement bug in bun that invalidates cache. Apparently this has to do with the custom @bunjavascript binary that ships with standalone Claude CLI.

The workaround there is to use Claude with `npx @anthropic-ai/claude-code`

2nd bug is worse, he claims that --resume always breaks cache. And there doesn't seem to be a workaround there, except pinning to a very old version (that will miss on tons of features)

This bug is also documented on Github and confirmed by other folks.

I won't entertain the conspiracy theories there that Anthropic "chooses" to ignore these bugs because it gets them more $$$, they are actively benefiting from everyone hitting as much cached tokens as possible, so this is absolutely a great find and it does align with my thoughts earlier.

The very sudden spike in reporting for this, the non-uniform nature (some folks are completely fine, some folks are hitting quotas after saying "hey") definitely points to a bug.

cc @trq212 @bcherny @_catwu for visibility in case this helps all of us.

[Eric Phy @EricPhy]

@theandreboso I upgraded to max5 about 2 months ago. This has happened at least 4 times since then.

It’s a recurring problem. Ppl will gaslight you and claim it’s you. It’s not; Claude usage goes crazy once every 1-2 weeks. Anthropic admitted it, and reset usage at least once already.

[Eric Phy @EricPhy]

@karpathy As someone who has argued with llms many times, this is incorrect. That said, they’re excellent at helping to refine existing arguments.

This sounds like you were just incorrect from the start. Which admittedly, is awesome that the llm helped you discover that.

[Eric Phy @EricPhy]

@karpathy Not to mention it’s often wrong. Need a good curated trash collector for llm memories.

[Linear @linear]

Issue tracking is dead.
We are building what comes next.

https://t.co/JM2d4mAEEB

[Eric Phy @EricPhy]

@linear Possibly your most based move yet. And necessary.