Alex Mitch

Solo founder building WordPress-native content software.
AI, agents, SEO, and the messy parts of shipping.

[Alex Mitch @Alek_Mitch]

Nailed it!

[Alex Mitch @Alek_Mitch]

Codex reset prediction update:

Damn, no reset yesterday and I was 68% sure. But my cumulative prediction stays the same:
Chance by end of Fri: 78%. That's today.

[Alex Mitch @Alek_Mitch]

@Gelassoldat @UnslothAI But even if you could, you'd still need some extra RAM for your system to keep running and other small things

[Alex Mitch @Alek_Mitch]

@kr0der Difference between Luna xHigh and Luna Max are huge. Still, Sol is Sol

[Alex Mitch @Alek_Mitch]

@thsottiaux @blueemi99

[Alex Mitch @Alek_Mitch]

@blueemi99 Was not the right time I presume. They will figure something out.

[Alex Mitch @Alek_Mitch]

@dieselville @Elger57005154 @thsottiaux @Robertg761_ why not use something like chrome remote desktop?

[Alex Mitch @Alek_Mitch]

@RightSideOfAI @astropol0 Silence on this question is nor surprisng.

[Alex Mitch @Alek_Mitch]

@SourceCodeplz @bridgebench @bridgemindai But how will they raise engagement on X then?

[Alex Mitch @Alek_Mitch]

@trikcode The next Deepseek v4 Pro (beta) will be =Sol at 'Deepseek pricing' lol

[Alex Mitch @Alek_Mitch]

@thsottiaux Appreciate making luna much more affordable in chatgpt auth sub. Now even the $20 subs can do meaningful work, and the 5x/20x even more so.

[Alex Mitch @Alek_Mitch]

@Simonsdad3 there are other options. There is also Luna over API. And on openrouter it's $0.1/0.6, very cheap, promotion.

[Alex Mitch @Alek_Mitch]

Codex reset prediction update:

Damn, no reset yesterday and I was 68% sure. But my cumulative prediction stays the same:
Chance by end of Fri: 78%. That's today.

[Alex Mitch @Alek_Mitch]

@Simonsdad3 There's still time.

[Alex Mitch @Alek_Mitch]

@TrueWysdoom @bridgemindai I think this is a troll account, or at least using troll strategies to increase engagement on X and grow this account, and nothing more. If we keep reacting to this, we lose. Best to spend our time on genuine users.

[Alex Mitch @Alek_Mitch]

@bridgemindai is this just a troll account? 😂

[Alex Mitch @Alek_Mitch]

@bridgemindai @deepseekailover Every deepseek model I've used at the time of release and afterwards had been exceptional for the time: Original, V3, R1, V4 Flash is used daily by many devs due to low pricing and decent performance. So where is this benchmaxxing coming from?

Oh that's right: From Anthropic.

[Alex Mitch @Alek_Mitch]

@bridgemindai You must have very low opinion of Deepseek.

[Alex Mitch @Alek_Mitch]

@shergilldotdev @thsottiaux Might be best to start using codex without counting on resets, and use other providers and models as supplemental (glm, kimi, cursor/composer, the new Deepseek flash 0731).

A reset has become a 'crutch'. It damages my workflow.

[Alex Mitch @Alek_Mitch]

@MrZhushijie @Harun11xx @deepseek_ai Exactly, they used Deepseek harness.

[Alex Mitch @Alek_Mitch]

@deepseek_ai It is already better than GLM 5.2, I wonder what V4 PRO would bring...

And this isn't even the final version of the Flash model, they label it as beta. The Pro version will be truly one of those 'Deepseek moments'.