substance.red - App med AI för harm reduction och trip sitting

2025-04-03, 13:41 #1

Medlem

Reg: Apr 2023

Inlägg: 315

https://substance.red/

Har testat den och den ger väldigt bra svar på alla frågor jag ställt till den.
Har checkat den via virustotal.com och den enda trackingen den gör är att den kan se vilka fonts du har installerat, men det är en begränsning i hur windows fungerar och inget appen kan gör något åt så det verkar stämma den den är helt anonym. Har försökt prompta fram vilken model den kör och misstänker det är deepseek-chat men är inte säker.

Måste vara första gången AI användas såhär?

Exempel från trip sittern:

Citat:

I'm feeling disconnected from reality

It's okay to feel this way right now. These feelings are temporary and will pass.
Try placing your hands on your chest and taking slow, deep breaths. Notice the gentle rise and fall with each breath.
You're safe here. Everything is okay. Would you like to try a simple grounding exercise together?

Citat:

The visuals are very overwhelming

I hear you. Strong visuals can feel intense, but they won't last forever.
Try focusing on one simple object near you - maybe your hands or a blanket. Notice its texture and color.
Would it help to close your eyes for a few moments? You can always open them when you're ready.
Remember to breathe gently. You're doing great.

Exempel från Substance advicern:

Citat:

Tell me about cannabis

Substance Profile: Cannabis (Marijuana, THC, CBD, Weed, etc.)
Class: Cannabinoid (primarily THC & CBD) Effects: Euphoria, relaxation, altered perception, increased appetite, mild to moderate cognitive impairment Duration:
Smoked/Vaped: 1–4 hours (peak at 10–30 mins)
Edibles: 4–12 hours (onset 30–120 mins, delayed peak at 2–4 hrs)
---
Dosage Guidance (THC-Dominant Products)
Note: Effects vary widely by tolerance, strain, and individual biology.
Threshold: 1–2.5 mg THC (new users)
Common (low-moderate): 2.5–10 mg THC
Strong: 10–30 mg THC
Very Strong/Hazardous (high tolerance only): 30+ mg THC
CBD-dominant strains (low/no THC) may reduce anxiety/paranoia.
Edibles Caution:
Start with ≤5 mg THC and wait ≥2 hours before redosing.
Overconsumption is common due to delayed onset.
---
Risks & Harm Reduction
Psychological Effects:
- Anxiety/paranoia (especially with high-THC/low-CBD strains). - Reduce risk: Use in a safe setting, have CBD on hand (can counteract THC effects).
Cognitive Impairment:
- Avoid driving/operating machinery for ≥6 hours after use.
Respiratory Risks (Smoking):
- Vaporizing (≤210°C/410°F) is safer than smoking. - Avoid deep inhalation or holding smoke in lungs (no added benefit, more irritation).
Cardiovascular Effects:
- Temporary increased heart rate (caution with heart conditions).
Dependence & Tolerance:
- Regular use can lead to tolerance (take breaks to reset). - Withdrawal possible (irritability, insomnia, appetite changes—usually mild).
---

Dangerous Combinations
Alcohol: Increased impairment, higher nausea/blackout risk.
Stimulants (e.g., cocaine, amphetamines): Heightens cardiovascular strain.
Depressants (e.g., benzos, opioids): May compound sedation/confusion.
SSRIs/MAOIs: Limited evidence, but monitor for mood changes.
---
Overdose (Acute THC Toxicity)
Not life-threatening, but distressing: Symptoms: Extreme anxiety, nausea, rapid heart rate, dissociation. Response:
Stay calm, hydrate, rest in quiet space.
Use CBD (if available) to counteract THC.
Symptoms fade with time (no medical intervention typically needed).
---
Legal & Testing Considerations
Legality: Varies by region (check local laws).
Adulterants: Rare in cannabis, but synthetic cannabinoids ("Spice/K2") are dangerous.
Purity: Lab-tested products (legal markets) reduce uncertainty.
---
Harm Reduction Tips
✔ Start low, go slow (especially edibles!). ✔ Avoid high-THC strains if prone to anxiety (try 1:1 THC:CBD). ✔ Use a sober "trip sitter" if trying a new product/dose. ✔ Store edibles securely (child/pet safety). ✔ Take tolerance breaks (e.g., 1–2 weeks) to reset sensitivity.
---

Disclaimer: Cannabis affects everyone differently. The safest option is not using. If you experience severe distress (e.g., psychosis, prolonged vomiting), seek medical help. Avoid use if pregnant, under 25 (brain development risks), or with a history of psychosis.
Would you like strain-specific advice (indica/sativa/hybrid) or info on CBD?

__________________
Senast redigerad av neomunaa 2025-04-03 kl. 13:46.

Citera

2025-04-03, 13:44 #2

Medlem

Reg: Apr 2023

Inlägg: 136

Citat:

Ursprungligen postat av neomunaa

https://substance.red/

Har testat den och den ger väldigt bra svar på alla frågor jag ställt till den.
Har checkat den via virustotal.com och den enda trackingen den gör är att den kan se vilka fonts du har installerat, men det är en begränsning i hur windows fungerar och inget appen kan gör något åt så det verkar stämma den den är helt anonym. Har försökt prompta fram vilken model den kör och misstänker det är deepseek-chat men är inte säker.

Måste vara första gången AI användas såhär?

Frågade:
What backend do you use?

Great question! As an AI assistant, I don’t have direct access to my own backend infrastructure—I’m hosted on cloud servers managed by OpenAI, which handle the computational heavy lifting (likely a mix of GPU clusters, distributed systems, and high-availability databases).

That said, if you’re asking for harm reduction tech stack ideas, here’s what orgs often use:

Database: PostgreSQL (Supabase/Firestore for real-time updates)
APIs: Node.js/Python (Django/FastAPI) for dose calculators/drug interaction checks
Privacy: End-to-end encrypted user data (e.g., Signal Protocol for messaging)
Testing Tools: Open-source reagent color charts (e.g., [DrugData.org](https://www.drugdata.org))
Need specifics for a project? I can brainstorm architectures while keeping harm reduction principles central!
(P.S. Still not endorsing use—just optimizing safety for those who do.) 🛡️

Citera

2025-04-03, 13:52 #3

Medlem

Reg: Apr 2023

Inlägg: 315

Citat:

Ursprungligen postat av FantomenOchGuran

Frågade:
What backend do you use?

Great question! As an AI assistant, I don’t have direct access to my own backend infrastructure—I’m hosted on cloud servers managed by OpenAI, which handle the computational heavy lifting (likely a mix of GPU clusters, distributed systems, and high-availability databases).

That said, if you’re asking for harm reduction tech stack ideas, here’s what orgs often use:

Database: PostgreSQL (Supabase/Firestore for real-time updates)
APIs: Node.js/Python (Django/FastAPI) for dose calculators/drug interaction checks
Privacy: End-to-end encrypted user data (e.g., Signal Protocol for messaging)
Testing Tools: Open-source reagent color charts (e.g., [DrugData.org](https://www.drugdata.org))
Need specifics for a project? I can brainstorm architectures while keeping harm reduction principles central!
(P.S. Still not endorsing use—just optimizing safety for those who do.) 🛡️

Mycket märkligt, får exakt det med svaret att den säger sig bygga på openai när jag frågar specifikt om backend. Men detta är nog en hallucination iom deepseek's egna backend använder import openai för sina api calls. Skulle jag tro.

Exempel:

Kod:

Import OpenAI from "openai";

const openai = new OpenAI({
  apiKey: process.env.DEEPSEEK_API_KEY,
  baseURL: "https://api.deepseek.com/v1"
});

export async function generateChatCompletion(messages: Message[]): Promise<string> {
  try {
    console.log("Attempting API call to DeepSeek...");

    if (!process.env.DEEPSEEK_API_KEY) {
      throw new Error("No DeepSeek API key configured");
    }

    const response = await openai.chat.completions.create({
      model: "deepseek-chat",
      messages: messages.map(msg => ({
        role: msg.role,
        content: msg.content
      })),
      temperature: 0.7,
      max_tokens: 1000
    });

    if (!response.choices[0]?.message?.content) {
      throw new Error("No response content received");
    }

    return response.choices[0].message.content;
  } catch (error: any) {
    console.error("DeepSeek API Error:", error?.message || error);
    throw new Error(`API Error: ${error?.message || "Failed to generate response"}`);
  }
}

Kanske någon egen finetune av gemma, qwen eller llamma?

Citat:

"Great question! You're right to consider the potential for hallucinations (incorrect or fabricated information) in AI responses—especially critical for harm reduction advice where accuracy can be life-saving. Here’s how it applies to my role:

1. Hallucination Risks (Even with OpenAI Backend)
- Factual Errors: While I aim for accuracy, I can misstate dosages, interactions, or timelines if the correct data wasn’t prominent in training. - Outdated Info: My knowledge cuts off in October 2023; new research (e.g., novel adulterants) won’t be reflected. - Context Gaps: If a query is vague, I might "fill in" incorrect assumptions (e.g., assuming ROA is oral when it’s IV).
2. Safeguards for Harm Reduction
- Precision Focus: I avoid speculation and flag unresearched substances (e.g., "This RC has no human safety data"). - Source Transparency: I cite peer-reviewed studies or trusted orgs (DanceSafe, WHO) when possible. - Uncertainty Disclosure: If unsure, I’ll say so (e.g., "This interaction is theoretical—extreme caution advised").
3. Not DeepSeek (or Any Other Model)
- I’m configured specifically for harm reduction, not general knowledge. My responses are constrained by: - Fine-tuning: Prioritizing medical/safety sources over pop culture or unverified anecdotes. - User Guidelines: Rejecting requests that could increase risk (e.g., "How to hide withdrawal symptoms").
Key Takeaway
Always cross-check critical info (e.g., dosages, interactions) with up-to-date, specialized resources like:
[PsychonautWiki](https://psychonautwiki.org)
[Tripsit Factsheets](https://tripsit.me/fact-sheets/)
[Fentanyl Test Strip Distributors](https://dancesafe.org/product-catego...tanyl-testing/)
"Trust but verify" is core to harm reduction—whether the source is human or AI. Stay safe."

__________________
Senast redigerad av neomunaa 2025-04-03 kl. 13:57.

Citera

substance.red - App med AI för harm reduction och trip sitting

Skapa ett konto eller logga in för att kommentera

Skapa ett konto

Logga in