Adi

That AI Guy. Focus- 1M leaders, orgs and communities transformed with AI. Love & Scared. serial entrepreneur, intellectual upstart, loves humor, reasonable.

[Adi @dearAdi]

@TheEconomist Love the British humour! good on you @economist !

[Adi @dearAdi]

@lydiahallie Instruction Following 😭

[Adi @dearAdi]

@kunchenguid Sigh agreed. It's negative instruction following, almost to the point where, at times, I thought it was maliciously sabotaging me.

[Adi @dearAdi]

@JeremyNguyenPhD Yeah no. Still needs extensive advisor / verification panels. But still quite a bit better than before.

[Adi @dearAdi]

@rohanpaul_ai Opus 5 as default with Sol Xtrahigh as advisor. Works really well.

[Adi @dearAdi]

@dabit3 Quick question: if you're getting a coding plan, can you just add the model OMP et al.? Docs are a bit unclear. Thanks!

[Adi @dearAdi]

@julien_c On the BBCs backup list "an online database" "a ai hobby website" "a social media site dedicated to platonic relationships" :)

[Adi @dearAdi]

@TheFP @bariweiss You rang?

[Adi @dearAdi]

@RenLeanna Welcome!

[Adi @dearAdi]

@elonmusk @BrianRoemmele @Not_the_Bee Agreed. Launching cool new thing soon based on this :).

[Adi @dearAdi]

Codex 5.6 Let's go.

[Adi @dearAdi]

Temporarily solved by getting Fable 5 to call on Codex 5.5 xhigh minions...

[Adi @dearAdi]

3rd plan down. Triaging token loss / leak. Clearly not sustainable.

[Adi @dearAdi]

3rd plan down. Triaging token loss / leak. Clearly not sustainable.

[Adi @dearAdi]

2nd Max account exhausted.
Loading next...
Fable 5 continues to impress!

[Adi @dearAdi]

2nd Max account exhausted.
Loading next...
Fable 5 continues to impress!

[Adi @dearAdi]

1 200 USD Max account done already.
Loading next....
Fable 5 goes hard.
@AnthropicAI

[Adi @dearAdi]

1 200 USD Max account done already.
Loading next....
Fable 5 goes hard.
@AnthropicAI

[Adi @dearAdi]

1 200 USD Max account done already.
Loading next....
Fable 5 goes hard.
@AnthropicAI

[Adi @dearAdi]

Fable.

[Adi @dearAdi]

@deredleritt3r Matches reasonably our internal evaluations for legal. For code of course would be incredibly different.
(GLM 5.2 is an incredible model)

[Adi @dearAdi]

@GoogleResearch Excellent. Where can we use it? Try it out? Not showing on the AI Studio Yet :(

[Adi @dearAdi]

@emollick Honestly, I still think about o1-pro.