AIvorax

Learning Al agents in public. Sharing what works, what breaks, and the prompts behind it

[AIvorax @AIvorax1]

@xsmryze I wonder why no one thought of this before, with this solution the tasks will be performed much better

[xsmryze @xsmryze]

THIS IS AN ACTUAL F*CKING GOLDMINE

Anthropic just shared one of the best tricks for Claude Code.

The idea: run Sonnet, Haiku 5.5 and Opus 5.5 at once, each one doing only the job it's best at.

Better results and way fewer tokens, because every task lands on the right model.

You just launch Claude Code like this:

claude --model sonnet --advisor opus

here's the split:

→ Sonnet 5.5 runs the session, writes the code and applies the fixes
→ Haiku 5.5 swarms the repo in parallel: search, grep and diff audits
→ Opus 5.5 stays on call as advisor for the big calls and signs off on the diff
→ you stop burning Opus context on work that doesn't need it

Opus only steps in at three moments:

→ before a plan locks
→ when the same error shows up twice
→ before calling it done

Sonnet builds. Haiku swarms. Opus reviews.

JEV engineering is the same move one layer down: the forks that need no thinker (which file, which tool, retry or stop) go to JEV in milliseconds, and the big models only see the ones that split

the full tree
> Sonnet 5.5 runs the main session and coordinates the work
> search maps the repo on Haiku 5.5, read-only
> grep finds call sites on Haiku 5.5 in its own context
> audit checks the diff on Haiku 5.5 in a throwaway worktree
> Opus 5.5 on call as advisor, reviews the synthesized diff before sign-off
> JEV routes micro-forks in milliseconds

paste the tree and this prompt into Claude Code ↓

"Reconfigure my Claude Code setup around this tree:
1. Run the main session on Sonnet 5.5 with Opus 5.5 as advisor.
2. Create three Haiku 5.5 subagents in .claude/agents: search (maps the repo), grep (finds call sites), audit (checks the diff).
3. Set isolation: worktree on audit so its checks never touch my working branch.
4. Add one rule to CLAUDE.md: delegate every repo search and test run to the Haiku 5.5 subagents to keep the main context small.
5. Route fast deterministic forks (tool choice, path selection, retries) through JEV.
Show me every change as a diff first. No edits until I say go."

↳ https://t.co/YqkjQiqRGv

save it, it'll come in handy.

[AIvorax @AIvorax1]

@Mr_Salio What people once had to create by hand and take weeks to do, can now be done in a few hours with code, it's just amazing

[AIvorax @AIvorax1]

@UP_SHIFTED In English, it is easier to convey your thoughts through a small selection of words, while in Ukrainian, the same word in context can mean completely different things. Therefore, AI has to analyze the context better

[AIvorax @AIvorax1]

@NodePrint I agree with you, in my opinion AI is not good enough to analyze people and identify the perpetrator, but it can still track him down

[AIvorax @AIvorax1]

OpenAI just said something that sounds like a movie plot.

Someone hid a recording device inside one of their offices.

So they pointed their own AI models at it. What is it, where did it come from, who put it there.

According to OpenAI, it traced back to one employee who allegedly planted it only a few days earlier. The case went to the FBI, and the person was arrested and fired.

They say there’s no sign that model weights, source code or customer data were touched. The investigation is still going.

A couple of years ago we were arguing about whether AI could write a decent email. Now it’s helping catch someone bugging an AI lab.

Weird times. Would you trust AI to investigate something like this?

[AIvorax @AIvorax1]

@IndependentEco AI is a good tool as long as it obeys you

[AIvorax @AIvorax1]

@UP_SHIFTED It seems that laboratories are trying their best to strengthen protection against hackers and therefore forget about such a simple danger

[AIvorax @AIvorax1]

@IndependentEco Well, this is just speculation, but honestly, a few years ago we couldn't even imagine what we would have today

[Eco @IndependentEco]

Opus 5.5 is throwing out model overloaded errors.

It looks like the finance bros have finally learned what Anthropic is.

The normies will now start to learn how Opus 5.5 can change your life.

We are still early but once Anthropic becomes the largest company in the world, everyone will be using it.

[AIvorax @AIvorax1]

@AI_cortexx I would be cautious about trusting such data to AI

[AIvorax @AIvorax1]

@renproton Now imagine the reactions of people who have been studying this for several years and now have lost their jobs

[AIvorax @AIvorax1]

@AInewssX I'm more interested in what changes will happen in the future

[AIvorax @AIvorax1]

Opus 5.5 is out, and the best part isn’t even a benchmark.

It just gets to the point. Answer first, explanation after. No four paragraphs of warm-up before the thing you actually asked for.

On top of that: ~40% cheaper than Opus 5, 30%+ faster output, and it holds up better on long multi-step agent tasks.

Smarter AND cheaper doesn’t happen often.

Anyone else switched yet?

[AInewssX @AInewssX]

Claude charges 20x less to reread text it has already seen. Most apps never turn it on.

Mark the repeated part of your prompt
The first request writes it to the cache
Every later request reads it at a discount

On Sonnet 5.5, a 20,000 token document asked 1,000 times:
Without caching: $40.00
With caching: $2.05

The catch: the first request costs more (1.25x), the cache lasts 5 minutes, and prompts under 512 tokens aren’t cached.

The hard part was never the discount. It’s remembering to turn it on ↓

Reply under the post (sources):
The $2.05 is my calculation from the public price list, not a benchmark.

The main post is about 450 characters, so it needs a longer-post account. If you need 280 or less, cut the catch paragraph and keep the rest.

[AIvorax @AIvorax1]

@UP_SHIFTED Now I'm waiting for the updated model, I wonder what else they'll come up with to make it better

[AIvorax @AIvorax1]

@semichenkko Opus never fails, I always choose it if I need to do something high-quality

[AIvorax @AIvorax1]

@UP_SHIFTED This will be an extremely important day for them, as their losses may be unjustified

[UP SHIFTED @UP_SHIFTED]

Four companies will spend over $700 billion on AI this year.

This week, we find out if it’s paying off.

Alphabet: $195–205B
Amazon: ~$200B
Microsoft: ~$190B
Meta: $135–145B
That’s about $2 billion a day. Almost double last year.
Microsoft and Alphabet report on Wednesday. Amazon on Thursday.
And Wall Street isn’t clapping for big numbers anymore. When Alphabet raised its AI budget by $15B in July, its stock fell 6% that day.

The demand is real. Google Cloud grew 82% last quarter.

The question is whether AI revenue can catch up with a $2-billion-a-day bill.

Wednesday is the first answer.