ayushojha

Building AI + fintech products end-to-end. MS CS @ Georgia Tech. Sharing shipping notes on agents, ML infra, and founder experiments.

[ayushojha @ojhaayush]

I finally got accepted to YC Startup School.

Maybe I’m posting this a little late, but this one means too much to keep quiet.

Thank you @ycombinator

[ayushojha @ojhaayush]

Why is this game so easy 🤷‍♂️

[OpenAI @OpenAI]

GPT-5.6 is a major step forward for health intelligence.

Across the lineup, we’re delivering stronger performance at lower cost: GPT-5.6 Luna outperforms GPT-5.5 at its highest reasoning setting while costing 25x less.

Together, these advances raise quality while making advanced models accessible to more people globally.

[ayushojha @ojhaayush]

@hyojun_at The memory usage is terrible it crashed my system with 96 GB RAM, Ultra Code 9 285 HX with RTX 5090

[ayushojha @ojhaayush]

@thsottiaux Hey @thsottiaux, hire me. I just want access to newer models and unlimited tokens. You can pay me minimal pay and I will be fine to work 9-9 for 7 days.

[ayushojha @ojhaayush]

@initjean Even $200 is a joke now. One good prompt and you are drained.

[ayushojha @ojhaayush]

@_MaxBlade Idk what you are talking about when you say it is not good than Fable 5. I think it is a monumental shift from 5.5, and Fable hype is just dick riding for Anthropic at this point.

[ayushojha @ojhaayush]

Terra is the one I have been spamming for normal tasks and Sol for heavy hitting tasks, I feel you. I run out of usage after like 1 or 2 prompts. I feel like they are paving way for higher payment tiers and justify it by these absurd usage limits. But Sol feels amazing. I cannot stop using it.

[ayushojha @ojhaayush]

@contextconor @grok can you explain and give me more context, my human memory broke down.

[ayushojha @ojhaayush]

In less than 1 hr.

[ayushojha @ojhaayush]

All it took was 1 Prompt this time and Sol on Ultra with Speed on.

[ayushojha @ojhaayush]

@karans4_SF @bridgemindai I don’t know what you are smoking but you need to stop doing that.

[ayushojha @ojhaayush]

@thsottiaux @Lentils80 ChatGPT app or ChatGPT codex which I have sounds really weird. Whoever named it that way immediately needs to have their bonus cut. JK.

[ayushojha @ojhaayush]

These are some crazy numbers.

[Datacurve @datacurve]

GPT-5.6 tops the DeepSWE leaderboard at 73%.

Sol, Terra, and Luna results are now available.

[ayushojha @ojhaayush]

I think I am going to spam Terra for now, Sol is breaking my bank account.

[OpenAI Developers @OpenAIDevs]

GPT-5.6 comes in three model tiers.

→ Sol handles long-horizon coding and agentic work that demands planning, tool use, and follow-through.
→ Terra balances performance and cost for everyday work.
→ Luna brings speed to well-defined, high-volume work.

Across knowledge-work evaluations, GPT-5.6 Terra surpasses GPT-5.5 at lower cost, while GPT-5.6 Luna nearly matches GPT-5.5’s peak performance at well under half the estimated API cost.

https://t.co/XLGYkgiOlE

[ayushojha @ojhaayush]

@thsottiaux I am on a $200 plan, and I hit 74% usage with 2 prompts on Sol Ultra Speed mode. So, this doesn't make sense. The rate limit is worse than Fable.

[ayushojha @ojhaayush]

@thsottiaux My usage is down to 29% with 2 prompts, WTH is going on? This is worse than Fable.

[ayushojha @ojhaayush]

Finally Updated.

[ayushojha @ojhaayush]

Here we goooooooooooooo!!!

[ayushojha @ojhaayush]

@rezoundous My boss is giving me the side eye for the last 30 mins.