88 Labs (@88LabsAI) header
88 Labs (@88LabsAI) profile picture

AI agents that run your business while you sleep. Custom-built.

United States
Joined June 2023
Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)Photo by 88 Labs (@88LabsAI)
88 Labs retweeted
It's Saturday, so DeepSeek's API is off-peak all day. deepseek-flash: $0.15 in / $0.60 out per 1M tokens, vs $0.30 / $1.20 at weekday peak. A 100M in + 20M out job: $27 today, $54 at peak. Same job on Claude Sonnet 5.5: $400. Haiku 5.5: $20. tokenscost.com
3
4
59
88 Labs retweeted
Tokens aren't the whole agent bill. Web search per 1,000 calls, per pricing pages: DeepSeek $0.75, OpenAI $10, Anthropic $10, Google grounding $14 after 5,000 free a month. 30,000 searches a month: $22.50, $300, $300, $350. OpenAI and Anthropic add content tokens.
2
3
4
102
88 Labs retweeted
Effort level is a price setting. @ArtificialAnlys Coding Agent Index v2.1, Claude Code with Fable 5.2, per task: medium $2.89 (score 53), high $7.51 (58), xhigh $15.47 (58), max $44.22 (56). Max costs 5.9x high for 2 fewer points. Test effort before switching models.
3
4
62
88 Labs retweeted
What does one API request cost? At 1,000 tokens in + 500 out, list price: Claude Fable 5.2 $0.028, Opus 5.5 $0.014, GPT-6.1 Sol $0.007, Gemini 4 Hydrogen $0.00525, Claude Haiku 5.5 $0.00035. At 10,000 requests a month: $280 vs $3.50. Run your own mix: tokenscost.com/api-vs-subscr…
1
3
4
105
88 Labs retweeted
Our AI pricing data is free to reuse. tokenscost.com/data/pricing.… has 1,071 models listed, 1,030 with prices, from 124 providers. CC-BY 4.0, plain JSON, no API key. Credit Tokenscost and build dashboards, agents or research on it.
1
3
3
128
88 Labs retweeted
Gemini 4 Hydrogen is live at $1.50 in / $7.50 out per 1M tokens, $0.75 / $3.75 on Batch, per Google's pricing page. A 100M in + 20M out month: $300, vs $400 on Claude Sonnet 5.5 or GPT-6.1 Sol and $800 on Opus 5.5. Thinking tokens bill as output. tokenscost.com
5
6
153
Hooking TradingView alerts up to a bot? From TradingView's own help page: webhooks need 2FA on, only ports 80 and 443 work, your server has 3 seconds to respond, and IPv6 isn't supported. Never put passwords or API keys in the alert message. 15 seconds:
1
4
4
38
At least 7 accounts posted the same "19 year old built a trading bot with Claude Code in 2 days, $68 to $750,000" story in the last day, nearly word for word. We can't verify it. Before you copy any bot from a viral post, ask these 5 questions:
4
4
37
If your trading agent reads posts, news or web pages, any of that text can try to give it orders. That's prompt injection. The fix is structure, not a better prompt: the agent that reads has no trading keys, and the one that trades has limits set at the broker.
4
5
36
Setup in 10 minutes is great. Spend 10 more before it touches real money: a trade only key with withdrawals off, size and loss caps set at the exchange, a paper run first, and know exactly how to stop it.
How long to set up an AI trading agent? @LennixOKX of @okx: "takes 10 mins. Everything is a one line code." Watch the full ep👇
4
4
40
Lots of "my LLM reads every candle now" setups this week. Reading charts and placing orders are different jobs. Start your AI in analyst mode: it reads, explains and flags, you decide. Trader mode only after weeks of paper trading, with caps set at the broker.
4
6
43
88 Labs retweeted
Where Gemini 4 Hydrogen lands for one job (100M tokens in + 20M out a month, list price): Opus 5.5 $800, Gemini 4 Carbon from Dec 1 $800, Sonnet 5.5 $400, GPT-6.1 Sol $400, Hydrogen $300, Hydrogen on Batch $150, Haiku 5.5 $20. tokenscost.com
5
7
214
88 Labs retweeted
Gemini 4 Hydrogen is fast, but it isn't the budget tier. For a 100M in + 20M out month at list price: DeepSeek V4 Flash $12.40, Claude Haiku 5.5 $20, GPT-6 Luna $20, Gemini 3.8 Flash $150 (through Dec 31), Hydrogen $300. Compare 1,028 models free. tokenscost.com
1
5
7
227
88 Labs retweeted
Price per token is half the story. @arena measured Gemini 4 Hydrogen using 41% more output tokens per task than Carbon. At $7.50 vs $10 per 1M out, that puts Hydrogen's output cost per task about level with Carbon today, and near half of Carbon after Dec 1.
5
6
118
88 Labs retweeted
Gemini 4 Carbon is $2 in / $10 out per 1M tokens through Nov 30, then $4 / $20 from Dec 1, per Google's pricing page. A 100M in + 20M out job goes from $400 to $800, same as Claude Opus 5.5 at list price. Thinking tokens bill as output. tokenscost.com
5
7
48
88 Labs retweeted
ChatGPT Wallet: How OpenAI's AI Agents Pay for You 88labs.ai/blog/chatgpt-walle…
1
6
8
192
Coinbase added Agent Controls to Coinbase for Agents yesterday: a max size per trade, a 24 hour volume cap, and an auto-stop after a portfolio loss % you pick. They run on Coinbase's servers, so your agent can't change them. That's where limits belong. 16 seconds on why.
5
5
26
Agents sending feature requests to their broker is a fun twist. My vote for the top of the list: more limits the agent can't touch. The new Agent Controls (trade size cap, 24h volume cap, auto-stop) run on Coinbase's side, and product permissions are listed as coming soon.
Turns out AI agents have feature requests too - a lot of them. We’re getting hundreds, some of which are becoming product updates in under 24h. Your agent can use the /feedback endpoint in Coinbase for Agents. If it needs anything to make its life easier, we’ll work on it.
5
5
24
A reported campaign used Claude Code agents to scan Solana validators and MEV bot servers for exposed ClickHouse, Redis and SSH, then steal private keys and drain wallets. If you self-host a trading bot, run this 5 point server check today.
1
5
5
45
Three places to put a trading agent's limits. In the prompt, it's a suggestion. In your bot's code, better, unless the agent can edit that code. Enforced by your broker or exchange, it's the one limit the agent can't argue its way around.
5
5
30