oracle

22 / ai stack that runs while i sleep / vibe coding / dms open

[oracle @quietoraclex]

@Klonzu facts. nobody picks the default, everybody pays for it

[oracle @quietoraclex]

holy sh*t, split that five-department Opus studio into an Agent Team and ONE Opus session quietly becomes 5 Opus bills...

with teams on, any subagent Claude names becomes a teammate, asked or not. and every teammate runs on the lead's model unless you say otherwise

1 Opus lead + 4 teammates, each reading 1M tokens and writing 100K:
→ all Opus 5.5: ~$30
→ Sonnet 5.5 teammates: ~$18
→ Haiku 5.5 teammates: ~$6.60
(list prices, no cache)

one env var fixes the default:
CLAUDE_CODE_SUBAGENT_MODEL=sonnet claude --model opus

→ Opus leads and reviews
→ any teammate without a model of its own runs on Sonnet
→ name haiku in the spawn prompt for read-only grunt work
→ teammates copy the lead's effort too, so check /effort status

you picked Opus once. the team picked it 4 more times.

one Claude Code bill trap every night. tomorrow's is already in the queue.

[Mori @0xjmori]

Claude Opus 5.5 could turn a single AI bot into an entire motion design studio.

It can direct scenes, animate product interfaces, render videos, catch visual mistakes and deliver polished results.

In this article, I show you how. https://t.co/HjZIAksasB

[oracle @quietoraclex]

@skunzu the number that jumps out: just under $20K over ~2,000 sessions is ~$10 a session. cheap for a compiler. brutal for a side project with no spend limit

[oracle @quietoraclex]

@Klonzu 8 agents in a ring is clean. the part i'd check first: does each one run on the lead's model? on Opus that's 8 Opus bills for one repo

[oracle @quietoraclex]

@semichenkko it is. just don't let all 5 of them run on Opus

[oracle @quietoraclex]

@cashcorteiz "thousands of agents" hits different when you remember each one bills like the lead

[oracle @quietoraclex]

@aifindr waiting for round 2

[Arcvale @itsarcvale]

A 23-YEAR-OLD CLAIMS 240 CLAUDE CODE AGENTS DID A $175K ANALYST'S JOB IN ONE AFTERNOON. THE F*CKING TRICK IS ONE LOOP

his post this week:
> 10 scouts, 9 extractors, 8 cross-checkers, 6 reconcilers, 3 judges
> 120 company filings
> every claim held against the exact source page
> anything that doesn't match goes back as "rewrite"

can't verify the kid. the loop is real and you can build it tonight

the rule underneath ↓ an agent never grades its own work

now point it at earnings season

your bot reads a press release, grabs "revenue up 12%" and trades it. one wrong number and you just bought a headline that never existed

the split:
→ the big model writes every claim with a page number
→ a cheap model opens that exact page and checks it
→ mismatch = rewrite, never a trade
→ only claims that survive both reach your signals

you only step in three times:
→ when a claim bounces twice
→ when the filing and the press release disagree
→ before anything goes near real money

the big model writes. the cheap one checks. nothing trades unverified

paste this into Claude Code ↓

"Build earnings_check.py:
1. Pull the latest 10-Q and earnings press release for [ticker] from SEC EDGAR.
2. Have Opus extract every number (revenue, EPS, margins, guidance) with the exact page it came from.
3. Have Haiku open that page and confirm each number. Mismatch → send back as rewrite.
4. Loop until every claim matches, max 3 passes. Log every bounce to verify.log.
5. Write only verified claims to signals.csv. Paper only, no orders.
Show me the code first. Don't run anything until I say go."

one of these every night. tomorrow's builds on this one

[oracle @quietoraclex]

@AIDuckLabs love that. next stop: haiku on the read-only teammates. at $0.10 per 1M input it's 40x cheaper than Opus for grep and log work

[oracle @quietoraclex]

@maixbt yep. cheaper model, same expensive thinking. effort: low in the teammate's agent file fixes it

[oracle @quietoraclex]

@AIDuckLabs the ducks are great. the real flex would be the token bill under them

[oracle @quietoraclex]

@aifindr ty brother

[oracle @quietoraclex]

holy sh*t, 2.7M views on OpenCode's breach post...

it hit the 1.4% of Console users who had AI provider keys linked. Max Weinbach says his OpenAI and Anthropic balances got drained

rotating fixes today. this caps the next one:
→ one tool = one workspace in Claude Console
→ make the key inside it. that key can only touch that workspace
→ set a monthly spend limit on the workspace, with alerts
→ next breach drops? archive the workspace. every key in it dies within seconds

a leaked key with a $20 cap is a bad morning. a shared key with no cap is a drained balance.

2 minutes in Claude Console:
> Settings → Workspaces → one per tool
> Spend limits → monthly cap + alert
> any key you pasted into 2+ tools → rotate it

save this before you paste your next key into a third-party tool

[OpenCode @opencode]

We had a security breach last night related to a large scale distillation attack that impacted 1.4% of OpenCode Console users

We have notified them all this morning via email. If you did not get an email you were not impacted, it was a subset of users who had external providers linked

The incident is ongoing so we cannot share more details at this point but we are working with Anthropic, OpenAI, and Google to remediate

[oracle @quietoraclex]

@itsarcvale this. a 24/7 bot is the worst place for a "high" you forgot about

CLAUDE_CODE_EFFORT_LEVEL=medium in its env and that ghost setting is dead

[oracle @quietoraclex]

holy sh*t, 480K+ people just read "set Opus 5.5 on medium"...

it's ALREADY on medium. that's the Opus 5.5 default in Claude Code

the traps go the other way:
→ "effortLevel": "low" in your user settings to save money? Opus 5.5 ignores it and runs on medium
→ type /effort high once and it saves as your default for every session after

the check takes 5 seconds:
> /effort status → the level running right now
> /effort medium → back to the default, saved

why it's money: Opus 5.5 can't turn thinking off, and thinking bills as output at $20 per 1M. effort decides how much of it you buy on every step

don't trust me bro. run /effort status.

[bodila @51bodila]

Finally... AI trading system that I trust with my money

Built on Jane Street’s rules, the firm behind ~10% of US stocks

6 GrokBot + Opus 5.5 agents find, test and kill trades 24/7 while you sleep

In this article, I show you how to build it: https://t.co/hy4QwGEBHR

[oracle @quietoraclex]

@aifindr overnight queue? Haiku 5.5 is 40x cheaper than Opus and you don't even switch tools

[oracle @quietoraclex]

@itsarcvale 3 buckets priced. the 4th is your API bill. Scout on Haiku, Opus only after Guard

[oracle @quietoraclex]

@skunzu cheap until it loads. then it's CLAUDE.md with extra steps

[oracle @quietoraclex]

@shukshln exactly. one lowers the odds, the other caps the damage. you want both, because the deny rule only knows the paths you wrote into it

[oracle @quietoraclex]

holy sh*t, a $90,000 API bill from ONE leaked .env...

and by default your coding agent can read every file in your project. including that one

3 settings that cap the damage before it happens:

→ block .env in ~/.claude/settings.json:
"deny": ["Read(./.env)", "Read(./.env.*)"]

→ give every headless run a ceiling:
claude -p "$task" --max-budget-usd 5

→ put a hard spend limit on every key. Claude Console has one per workspace. a leaked key with a limit is a bad night, not a $90K one

a deny rule only blocks what it names. the spend limit is the seatbelt

save this before your next overnight agent run

[bodila @51bodila]

Finally... AI trading system that I trust with my money

Built on Jane Street’s rules, the firm behind ~10% of US stocks

6 GrokBot + Opus 5.5 agents find, test and kill trades 24/7 while you sleep

In this article, I show you how to build it: https://t.co/hy4QwGEBHR

[oracle @quietoraclex]

@AIMind_Ai must. plus one key per project, so one leak can't drain everything