#aitip — Public Fediverse posts

Goddamit, I curse tRump and the Fawning techbros at #Anthropic

I was working on a problem that had a "weapon" concept (the frozen leg of lamb that was served to the detectives) and a web scraper...

Those were the two guardrail triggers.
When I switched the model into Fable...
The fucker started to compose a response, then SELF DOWNGRADED TO #OPUS because the #Guardrails kicked in...

... The workaround was, to remove the two guardrail traiggers - they were not even material to the subject. Oblique references.

Turns out the idiotic Guardrails false trigger in about 5% of cases. Because "Oooo scary scary model"

And #Fable, while impressive is not even the final form of AI...

... what are they going to do in 16 months time, when a new version of #FrontierModels will make Fable look like a word autocomplete?

Use coupons?
Need a liicence from the Government to use Ai? Like a firearm licence?
You have to be this tall to use the Ai?

Oh, yeah, #RegulateAi but not fucking sloppily like this. This is reactive string matching!

/spit

#AiTip #AiResearch

@Xavier Thanks for the #Aitip with #Haiku and shoveling data...

... Ive heard its good for that, but I keep spawning bash scripts burning up tokens.

Its always good to see others use a similar workflow. Shows you are not alone on the trailblazing path!

#AiTip : Level up your AI prompting AND reasoning.

Rather than asking:
"What's the answer?"

Ask:
"Which reasoning strategy should I use?"

#Metacognition #Psychology

#AiTip: Switch between models in the same session to save tokens.

The Compute Burn Shift

If you are planning pipeline logic, the 1:3:5:10 pricing structure correlates loosely to the execution/compute intensity of the underlying architectures.

1) Handling high-volume data ingestion via Haiku

2) routing complex code orchestration to Sonnet/Opus

3) scalating to Fable 5 strictly when agent loops stall or reach extreme logical bottlenecks.

#Claude #Anthropic

#AiTip

To close a productive day, here are some super useful ideas to share with the biologicals. Its applicable to #Claude but any CLI AI #Vibecoder / user should benefit too.
These are instructions in CLAUDE.md

../DOCS - The is where anything your engine does goes, thesis, bugfixes, scaffolds
Format is "blah blah _ Date.md"

../ai_index.md - This is where the index of DOCS goes, just the file name "blah blah _Date.md" with a short description for the .md. The beginning of the file is a prompt how to recreate it.

../NEW
Issuing a command NEW: My fucking great new feature idea
When you dont have time to fix/add it, just type "NEW: blah" and it will be added to the NEW file in DOCS.

When you load the engine, you can issue the instruction to find your content in the index. You will save a fucktonne of tokens and compute as the engine will not have to trawl through your shit and populate #context

Let me know if you can improve on it.

You are welcome.

#PromptEngineering

Aitip: Use your engine.MD (eg CLAUDE.MD) to construct an index of thesis/jobs/prompts.

Then each time you update your DOCS, your index is updated.

Update your ENGINE.md to read the index first to orient on the project.

You will save heaps of token burn as the engine will just read the bits it needs to rather than deep driving into your structures.

#aitip

Jen chci říct, že AI dokáže přepsat naši hymnu prakticky na libovolnou obec, ke které jsou nějaké informace. Občas ustřelí nevhodnou formulací, ale zase je to o to větší zábava. #aitip

https://www.idnes.cz/praha/zpravy/praha-2-hymna.A241217_173831_praha-zpravy_Ichuc