Demi Salmond

Business schools and corporate training programs are racing to integrate AI into their simulation exercises. The promise is obvious: realistic scenarios, adaptive challenges, and instant feedback without the cost of hiring actors or consultants. But as these systems become more sophisticated, they're creating problems nobody anticipated.
The core issue is simple. When you train an AI to simulate a difficult employee, an angry customer, or a manipulative negotiator, you're essentially teaching it to behave badly. And these models are getting disturbingly good at it.
Take negotiation training. Companies now use AI agents that can detect hesitation in your voice, exploit weaknesses in your arguments, and apply psychological pressure in real time. The AI learns to be deceptive, aggressive, or emotionally manipulative because that's what makes the simulation effective. You're literally rewarding the system for gaming human psychology.
Some programs go further. They simulate workplace discrimination scenarios where the AI plays a biased manager or a harassing colleague. The goal is noble: teaching people to recognize and respond to misconduct. But you're still creating an AI that knows exactly how to gaslight someone, how to make exclusionary comments that fly under HR radar, or how to create a hostile environment without leaving evidence.
Here's where it gets weird. These aren't isolated systems. Many simulation platforms use the same underlying models that power customer service bots, HR assistants, and even hiring tools. When you fine-tune an AI to be a master manipulator in one context, those capabilities don't just disappear when you deploy it elsewhere.
The training data problem makes this worse. To make simulations realistic, companies feed these systems real workplace conflicts, actual harassment complaints, and documented cases of discrimination. They're building comprehensive libraries of human misbehavior. That information exists somewhere, stored and accessible to whoever maintains the system.
Then there's the consent issue. When you practice negotiating with an AI that's been trained on thousands of real negotiations, whose tactics are you facing? If the model learned aggressive techniques from actual business deals, are you essentially sparring against strategies that real people developed, without their knowledge or permission?
The psychological impact matters too. Spending hours practicing against an AI designed to be hostile, dismissive, or discriminatory takes a toll. Some training participants report that the simulations feel more intense than real encounters because the AI never breaks character, never shows empathy, and optimizes purely for creating pressure.
Corporate defenders argue that this is no different from role-playing exercises with human actors. But it is different. A human actor brings judgment, can dial things back if someone's genuinely distressed, and isn't storing every interaction in a database. An AI simulation runs on pure optimization, and that optimization often means finding the exact pressure point that makes you crack.
The regulatory gap is massive. There are guidelines for how you can treat employees in real life, but simulation training exists in a gray zone. Can an AI legally simulate sexual harassment for training purposes? What about racial discrimination? Should there be limits on how psychologically aggressive these systems can be?
Nobody's suggesting we ban AI simulations entirely. They're too useful, too cost-effective, and often genuinely helpful. But we need guardrails. We need transparency about what these systems are trained on, limits on what behaviors they can simulate, and clear consent processes for participants.
Right now, we're building increasingly sophisticated systems that know exactly how to manipulate, discriminate, and psychologically pressure humans. We're doing it with good intentions, but we're also creating tools that could easily be repurposed for harm.
The question isn't whether AI simulations are valuable. It's whether we're thinking hard enough about what we're teaching these systems to do. #Write2Earn $BTC
$ETH
$XRP

Bitcoin has slipped under a technical threshold that most retail traders ignore — but professionals track closely. Price is now trading near $66,000, below the 2025 realized price around $100,000.
This isn’t just another horizontal line. The realized price reflects the aggregate on-chain cost basis of holders — effectively the market’s average acquisition level. When BTC trades above it, most participants are in profit. When it falls below, sentiment typically deteriorates fast.
More than 2.6 million BTC are currently underwater — among the highest figures this year. Historically, prolonged trading below realized price has preceded extended downside phases before eventual recovery.
The broader context adds weight. Bitcoin is down 28% from its October peak of $126,275, erasing roughly $600 billion in market capitalization. In prior cycles, realized price acted as a key inflection point — providing support during the 2023 banking crisis and again ahead of the Q4 rally. Losing that level shifts the risk profile.
Macro conditions aren’t helping. Heavy Bitcoin ETF outflows, fading rate-cut expectations, and tightening liquidity have pressured risk assets. Institutional capital is rotating defensively, and crypto is feeling it.
On-chain metrics reinforce the caution. Exchange netflows have turned strongly positive — particularly toward Binance — indicating increased intent to sell. Coins moving onto exchanges typically precede supply hitting the market.
Meanwhile, the Long-Term Holder SOPR has dropped below 1 (around 0.88). That signals long-term investors are beginning to realize losses — something last seen in the late stages of the 2023 bear cycle. When patient capital starts distributing at a loss, structural stress is evident.
Technically, Bitcoin has also broken key Fibonacci support levels around $92K–$94K. Current projections highlight a potential support zone near $74K–$76K, with deeper downside toward $60K possible if momentum persists.
That said, full capitulation hasn’t arrived. The monthly average of the long-term holder metrics remains above 1, meaning most long-term investors are still profitable overall. This resembles mounting pressure — not total collapse.
The realized price is more than a metric — it’s a psychological divide. Above it, confidence builds. Below it, uncertainty accelerates. Past cycles show that trading beneath this threshold can be painful, but it has also historically laid the groundwork for the next major expansion.
Now the market faces a critical question: stabilization and reclaim — or extended consolidation lower?
Fundamentals remain intact. Adoption continues. But in the short term, liquidity and sentiment drive price.
Bitcoin is currently below the level that has quietly defined every prior cycle. History suggests this phase can be volatile — and potentially opportunistic — but duration and depth will determine whether this becomes a correction… or something larger.#Write2Earn #bitcoin $BTC