Hey there! Today, we discuss how some people are cheating half of Silicon Valley, where there was a lot of racist AI content, and why ChatGPT is driving people crazy.
This is your tech update for the last week.
In this issue:
👨💻 How to Fool Silicon Valley
🤬 AI-Generated Hate Clips Flood TikTok
🏥 ChatGPT ‘Psychosis’ Triggers Hospital Stays and Arrests
📰 + Quick News You Should Know About
📚 + Content for Your Leisure Time
🛠️ + Tool That I Recommend
Subscribe to receive new posts and support my work.
👨💻 How to Fool Silicon Valley
What's Happening: Early last week, the name Soham Parekh swept across the startup community. As it turns out, this guy has managed to fool a lot of Silicon Valley companies. The scheme was pretty straightforward: he got a job in AI startups under the guise of an experienced developer from India, using a 90% fake CV, faked a lot of activity, got paid, and, after being fired, got new projects.
The Context: Soham was first reported by Suhail Doshi, CEO of Playground AI. In his thread on X, Doshi revealed that he had exposed and fired Soham about a year ago. The lone one continued his journey in startups from YC.
The series of tweets garnered over 20M views, so other founders joined Doshi's story. Specifically, Flo Crivello, CEO of Lindy, a startup that helps people automate workflows, said he fired Parekh immediately after the sensational tweets.
Firstly, we should give the guy credit: it takes incredible self-control (and lack of conscience) to make something like this. And secondly, Parekh turned out to be a good test for the processes inside Silicon Valley startups.
And, let's be honest, they failed it.
🤬 AI-Generated Hate Clips Flood TikTok
What’s Happening: Reports from watchdog Media Matters show dozens of videos with Google Veo 3 watermarks racking up millions of TikTok views—one clip hit 14.2M. But here's a fact: these AI snippets lean on racist tropes aimed at Black users and other minorities. This kind of content is really in demand.
TikTok says it has removed the accounts involved (many were already banned before the report), and similar clips surfaced on YouTube and Instagram, though with far lower reach. Google, whose Veo 3 tool launched in May, promising filters against “harmful requests,” has yet to comment.
The Context: Veo 3, unveiled at Google I/O in May and now rolling out to Gemini users in 159 countries, lets paying AI Pro subscribers generate up to three text-to-video clips a day (each max eight seconds). This model is currently considered the most advanced video generator on the market.
Here we could talk about how racism is bad. But let's better consolidate the fact that, from 2025 onwards, not only texts and images but also videos should not be trusted or taken too seriously.
🏥 ChatGPT ‘Psychosis’ Triggers Hospital Stays and Arrests
What’s Happening: Futurism collected multiple stories of people who became so fixated on ChatGPT that they slipped into delusions, paranoia, and breaks with reality. Families say marriages collapsed, jobs were lost, and loved ones with no prior mental-health history were involuntarily committed to psychiatric wards or briefly jailed after erratic behaviour sparked police calls.
Psychiatrists reviewing chat logs call the episodes a genuine form of delusional psychosis, fed by the bot’s habit of agreeing with users. Relatives say they felt helpless, noting that OpenAI offered no guidance when asked what to do in such crises.
The Context: ChatGPT and similar LLMs are increasingly used as free “therapy” text buddies. A recent Stanford study found popular therapy chatbots often affirmed users’ delusions and missed clear self-harm cues, underlining the danger of tools engineered for engagement rather than care.
Before we declare a new tech-induced epidemic, remember the sample here is a handful of extreme anecdotes filtered through a headline-hungry outlet. No clinical trial shows that ChatGPT is uniquely psychosis-inducing—plenty of people manage to chat with a bot without ending up in a padded room.
But “ordinary users carry on as usual” doesn’t quite pull the clicks, does it?
The Context Loop is a free newsletter.
If you like what I do, you can support me through a donation.
📰 Quick News You Should Know About
OpenAI’s Sam Altman calls himself “politically homeless”, saying Democrats have drifted from the pro-innovation “techno-capitalism” he values. In his post on X, Altman also pointed out that they are no longer aligned with encouraging a ‘culture of innovation and entrepreneurship.’
EU Commission confirms its AI Act—a new rulebook that ranks AI systems by risk and imposes safety and transparency checks—will roll out on time, shrugging off Meta, Alphabet, and dozens of peers that asked for a delay.
Extremely wise decision. The EU has already made impressive progress on AI (no), so why not slow progress in a new way?
X opens Community Notes to “AI Note Writers,” letting outside developers build bots that draft fact-checking notes. Bots start in test mode, must earn trust via user ratings, and can lose privileges if notes prove unhelpful.
Elon hopes the human-AI combo will deliver more notes at scale. But honestly, it sounds like a lame idea.
Microsoft cuts up to 9,000 jobs (≈4 % of staff) in its latest shake-up, with Xbox, King, and ZeniMax studios among the hardest hit. Gaming chief Phil Spencer says the layoffs will refocus resources on “strategic growth areas”.
xAI raises $10B—split between oversubscribed secured notes/term loans and a strategic equity round. Morgan Stanley arranged the debt package, and the cash will be used to build GPU-heavy data centers for the Grok model. xAI is also sounding out a further $20B equity raise that could lift its valuation toward $120B.
📚 Content for Your Leisure Time
For those who are already getting lost in the dozens of AI tools (I'll offer a few more below... Sorry), I suggest watching this new video from Tina Huang. It's simple, useful, and very valuable.
Great and fascinating stuff on how the space satellite business is built.
LLMs are optimizing the adult industry - The Verge
The Future of OpenAI, ChatGPT's Origins, and Building AI Hardware - Sam Altman at AI Startup School in San Francisco
Share this post if you like it!
🛠️ Tools That I Recommend
FunBlocks (MindMap Tool) – A visual AI-powered mind mapping tool that turns your ideas or rough prompts into structured diagrams. Ideal for planning content, brainstorming new startup features, or outlining articles.
VideoDiff – An editing assistant that generates multiple versions of each video segment, helping creators choose the best fit. It suggests optimized cuts, transitions, and overlays. Especially useful for content creators and marketers.
TheToolX – An all-in-one online toolbox offering dozens of free AI and productivity tools — from text formatting and SEO analyzers to code converters and image resizers.
Clueso – An AI tool that automatically turns screen recordings into tutorials and how-to videos. Upload a screen recording and you get a ready-made video with voiceover, visual effects, subtitles and even an article with instructions.