@future.with.ai98: A lot of people think Claude got nerfed 🤯 But there’s a fix. If Claude Code feels weaker, slower, or less intelligent, you can tweak the reasoning settings inside the config.json file and instantly make it far more powerful again. One small setting change can completely transform how Claude performs. #Claude #ClaudeCode #AI #ArtificialIntelligence #Coding
Claude told me that anyone who shouts a zillion is a 🔔 end.
2026-04-15 19:13:24
0
reed :
i like the idea of low-code, but sometimes it still feels like you need to know too much. been using workbeaver lately and it’s been a smoother experience for me. i spend less time figuring things out and more time actually getting work done
2026-04-17 17:16:53
62
distractedmind007 :
Practical analogy: Think of it like configuring a contractor — instead of letting them decide how much effort to put in and which tools to use, you're locking in "always use the good tools, always work at full effort."
2026-04-16 23:19:52
0
Dodgey Dodgey :
is it 1 zillion or 1 thousand times better, you have to be sure it's very important
2026-04-15 16:03:11
5
Z_him_self :
Why do you have to begin every of your videos with a loud exaggeration
2026-04-15 17:07:29
0
Léo Beraldi :
Good tip, but worth some caveats:
What this actually does: disables newer adaptive features (1M context, adaptive thinking, auto memory) to get more consistent behavior, and routes subagents to Sonnet to cut costs. The effortLevel: high counterbalances by making the main model think harder.
The “nerfed” framing is subjective — some of these features genuinely add variability, so disabling them can feel like a restoration even if no regression happened.
Bigger concern: several of these env vars (CLAUDE_CODE_DISABLE_1M_CONTEXT, DISABLE_ADAPTIVE_THINKING) are not in Anthropic’s official docs. No public evidence they do what the names suggest, and no guarantee they survive the next update. You could be adding config that does nothing — or breaks silently later.
The one clearly documented and useful setting here is CLAUDE_CODE_SUBAGENT_MODEL: "sonnet" for cost control.
Test before committing — don’t take undocumented env vars at face value.
2026-05-25 23:43:01
0
Kevin | Building Tale Forge :
…. So what does the change actually do?
2026-04-15 16:55:31
0
Dreams by machine :
Can you tell me how to get it?
2026-05-26 06:06:35
0
Maestro19 :
what method, what settings to change...details please. reasoning..
2026-04-15 16:07:33
4
ForwardCompatible :
it worked until this morning when they rolled Opus 4.7. now it's hit or miss. It seems to still work with Sonnet 4.6
2026-04-16 19:24:42
0
Matrix guy :
Go to codex and it’s fixed
2026-04-15 21:29:39
1
Andrew Bryce :
I think he’s deleted the post as it didn’t work 😅
2026-04-15 12:18:49
1
Not Likely :
the best model is now Opus not sonnet.
2026-04-18 06:08:37
0
ChrisB :
And how quickly does it burn through tokens and hit it's daily limits?
2026-05-03 05:39:59
0
Shaunie :
all these made-up stats
2026-04-15 20:57:58
0
Kachorek Christian :
So… AGI?
2026-04-15 14:20:49
0
Daniel Sam :
Yes!! It’s back!!
2026-04-15 15:13:04
0
To see more videos from user @future.with.ai98, please go to the Tikwm
homepage.