nobody tells you this about grok bot
one guy spent a week reading everything about @bot: 600+ posts, 300 reels, 60 hours of video, the docs.
what he found:
> all your bots share ONE computer - same files, same cookies, same logins
> a 15-min routine = 96 runs a day. one hourly routine "can use a week of usage in a day"
> the terminal is a back door: type claude, log in your own Claude Code, drive it from your phone
plus 3 things xAI hid on purpose.
the full list, with every source, is in his article below 👇
Claude Code tip: once Opus 5.5 is your main model, stop letting your Fable 5.1 quota go to waste
put it on call with /advisor
run /advisor fable
Opus 5.5 does all the building
Fable 5.1 follows the session in the background and cuts in three times:
-> before a plan: does this hold up?
-> when one error keeps returning: am I fixing the wrong thing?
-> before "done": what slipped through?
Fable 5.1 checks the direction. Opus 5.5 does the typing
Jev engineering takes it one step further:
the small forks (which file, which tool, retry or stop) go to Jev in under half a second, so the pricey models only see the calls that can change the result
where this setup pays off most right now: long reverse engineering runs with REA, the open source MCP that lets Opus 5.5 take any app apart down to the binary
• the full setup
> Opus 5.5 on high drives the main session
> subagent one reads the code
> subagent two edits and runs tests
> subagent three checks the docs
> all three on medium
> Fable 5.1 on call
give Claude Code the tree plus this prompt 👇
"Configure Claude Code to match this tree:
1. Reuse suitable subagents from ~/.claude/agents and .claude/agents.
> Draft new ones only where a role is missing
> Give each model: opus, effort: medium
> List any that pin another model and leave them
2. Set main session effort to high with effortLevel in ~/.claude/settings.json
3. Look for env vars that turn off the advisor (CLAUDE_CODE_DISABLE_ADVISOR_TOOL, DISABLE_TELEMETRY, anything that blocks flag fetching) and CLAUDE_CODE_EFFORT_LEVEL, which overrides subagent effort. Report them and leave them be
4. Add a rule to ~/.claude/CLAUDE.md: check with the advisor before a big plan, when an error repeats, and before marking a long task done
Show me the changes first, edit nothing until I confirm."
↳ https://t.co/pkgl3RZv47
i just took apart photoshop with REA + opus 5.5... i was paying $275.88 a year for 4 buttons. now they live in a 212 KB app
REA takes it apart. opus 5.5 explains only my 4 buttons. claude code rebuilds them. i make thumbnails
photoshop → REA maps every function → opus 5.5 reads only my 4 tools → claude code rebuilds → 40 test images against the original
REA found 14,862 functions inside photoshop. opus 5.5 doesn't touch most of them. it only gets the 4 tools i use on every thumbnail
remove background, crop to 1280x720, curves, export under 2 MB. that's it. that's my whole photoshop
those 4 tools sit on 286 functions, read in 31 minutes for $9.73. 148 of them are behind 1 button - remove background
reading those by hand at 25 minutes each is 119h 10m. people would say 100x faster. i did the math and it's 231x
that's $0.034 per function, and every answer says what the button does, where the proof is, and what's still a guess
last night claude code rebuilt all 4 in 3h 12m. i ran 40 old thumbnails through both. 37 out of 40 came out the same
the other 3 had worse hair edges. my app opens in 0.8 seconds, photoshop took 14
the big mistake: giving opus 5.5 the whole photoshop. you pay the smartest model to read 14,862 functions and say "that's a 3d menu" 14,576 times
it doesn't rebuild photoshop. it rebuilds my 4 buttons. for anything else i'd still open the real one
setup took me 11 minutes: one REA command, photoshop already on my pc, opus 5.5 inside claude code, and 40 old thumbnails as tests
REA maps, opus 5.5 explains, claude code rebuilds what the button does - never a line of adobe's code
everyone pays adobe for 1,000 tools. i rebuilt the 4 i use - $22.99 a month, $275.88 a year back ↓
i just built an army of 500 Haiku 5.5 agents and made Opus 5.5 the general
Anthropic dropped Haiku 5.5 yesterday. it costs 75% less than Haiku 4.5
everyone saw a cheaper chatbot. i saw cheap soldiers
so i split the roles:
Opus 5.5 = the general. plans, decides, checks
Haiku 5.5 = the army. reads, sorts, summarizes
1 task → Opus cuts it into 500 pieces → 500 Haiku run at once → Opus merges the answer
first mission: read 10,000 pages of SEC filings from NVIDIA, Tesla, Apple, Microsoft and Google
a Goldman Sachs analyst: 3 weeks
Opus 5.5 alone: 6 hours, $412
my army: 11 minutes, $9
i did the math. 45x cheaper. 32x faster
the big mistake: people make Opus do everything
that's like paying Jensen Huang to sort your mail
setup took 14 minutes in Claude Code. one prompt, one subagent file
OpenAI and Google are fighting to build the smartest model
the money is in the cheapest one that never gets tired
also i do same thing with Grok + Jev, check it out in article below ↓