fedefring

[fedefring @fedefring]

@thsottiaux Same daily task, same question: "suggest an appropriate model and level of reasoning." Sol 5.6: "Terra High is suitable for the task." Sol 6: "set Astra Extra-High." No comment. Sure, really, Sol 6 is a step forward?

[fedefring @fedefring]

@thsottiaux Same daily task, same question: "suggest an appropriate model and level of reasoning." Sol 5.6: "Terra High is suitable for the task." Sol 6: "set Astra Extra-High." No comment. Sure, really, Sol 6 is a step forward?

[fedefring @fedefring]

@thsottiaux "Yes: you correctly described my mistake.
This does not demonstrate that GPT-6 Sol is less capable than 5.6 Sol. It demonstrates that I applied the gate incorrectly." No comment. 5h full session lost.

[fedefring @fedefring]

@thsottiaux "Yes: you correctly described my mistake.
This does not demonstrate that GPT-6 Sol is less capable than 5.6 Sol. It demonstrates that I applied the gate incorrectly." No comment. 5h full session lost.

[fedefring @fedefring]

@thsottiaux We’ll see. Same huge hype and expectation for Sol. And still struggling. We’ll see.

[fedefring @fedefring]

@thsottiaux don't know what to say. Are you really really sure nothing changed ??

[fedefring @fedefring]

@thsottiaux Ehm... sure nothing changed? today 100% first slot (within 5 hours limit) gone with less than 30 MIN working time mainly Terra High.. Sol one only time.. Codex was not even able to make a test of the script, it stucked in the middle, cause of pop-up limit. NC

[fedefring @fedefring]

@thsottiaux Guys, you are awesome. Never enough say "Thank you". Best

[fedefring @fedefring]

@thsottiaux A huge thank you !!!

[fedefring @fedefring]

@thsottiaux @grok
what date and time for Italy?

[fedefring @fedefring]

@thsottiaux Right choice friends. Less new models, but really no-bugs, efficient and reliable stable models. This is the way guys.

[fedefring @fedefring]

@thsottiaux Investigation hint: Codex burned ~4% of my "Weekly" usage getting stuck on 3 orchestration issues: 1) ghost subagents, 2) unified-diff deletion lines when content started with “--”, 3) PowerShell JavaScript string quoting. All were non-kernel issues>by Chat

[fedefring @fedefring]

@YashasGunderia @thsottiaux same

[fedefring @fedefring]

@thsottiaux Exactly same as @jamesfindsout here. Insane. Please solve this Major issue. No sub2api at all. Said to say and hard to do, but I am forced to look for other AI provider now. This way is completely insane.

[fedefring @fedefring]

@thsottiaux "Sol Ultra" also makes and corrects many mistakes. Let it correct, understand, and "block"/prevent wasting tokens/limits/money. Also, consider adding automatic switching between models to optimize consumption... always under supervision.

[fedefring @fedefring]

@thsottiaux

[fedefring @fedefring]

With Sol 5.6, one person was able to think, build, test, diagnose, correct, optimize, and document like an entire elite Linux engineering organization.

A project that should have demanded a company was completed by one individual in only a few weeks.

Only Sol 5.6 can.

[fedefring @fedefring]

In just a few weeks, one person accomplished what would normally require an entire company of senior Linux developers working for months.

Faced with a particular device running a particular version of embedded Linux, every obstacle seemed final:

[fedefring @fedefring]

Without Sol 5.6, work of this scale would have required multiple departments, specialist teams, endless meetings, complex coordination, and months upon months of senior engineering effort.

[fedefring @fedefring]

This was not merely software development. It was systems engineering, reverse engineering, experimental research, kernel development, driver debugging, automation, forensic analysis, and rigorous technical documentation, all concentrated in the hands of a single person.

[fedefring @fedefring]

Every anomaly became a testable hypothesis. Every reboot became a controlled experiment. Logs, traces, configurations, timings, and measurements were correlated until they revealed behaviors that no official documentation described with sufficient depth or operational clarity.