Sponsored by
✦
Friday debrief, people. 🕶️
So this week, the platform everybody trusted to stay neutral caught an AI jailbreak and then maybe sold itself to the chipmaker it swore off. Perimeter defense is having a week.
Meanwhile, the bigger question rattling around the stack: if you can't trust the exits, the models, or the platforms, what exactly are you hardening your dependencies against? The answer involves fewer checkboxes than you'd hope.
✦
The founder’s dashboard / Your quick roadmap
✦
FOUNDER BOARD
🤖 $0 virtual exec team built by fired devs puts AI in the C-suite, try a virtual exec now.
🛠️ 5 moves shift you from control to systems so teams run without you stop being the bottleneck this quarter.
🌅 First 60 minutes decide your output, lock in deep work hours.
💬 1 runway talk before equity prevents silent deadline gaps and mismatched risk, set salary-free timelines now
🧾 Creator equity now needs 5 vetted roles in writing, paper the deliverables before any grant.
You've built a company that runs without you — which move was the hardest for you to actually make?
Startup Investor Finder
Filter by stage, industry, and location, organize outreach, and manage your fundraising pipeline in one streamlined workspace. No random spreadsheets, no scattered LinkedIn tabs; just a focused system to turn research into real fundraising traction.
✦
RUSHIN' ROULETTE
Five bullets of updates
🤖 Viral AI startup Instinct raises $350 million at a $2.5 billion valuation, just one year after launch. The funding cements investor appetite for fast-scaling consumer AI even as regulators and users scrutinize privacy risks, signaling that growth is outpacing guardrails.
🌊 Seaglider maker Regentcraft is raising $240M in Series B funding to scale manufacturing of its coastal-hopping electric vessels. The round highlights investor appetite for dual-use hardware that can serve both defense and commercial transport, rewarding startups that can turn novel vehicles into certifiable, factory-built products.
🗺️ Former California utility PG&E engineer's mapping startup raises $26 million Series A to build 'Google Maps' for underground utilities. The funding backs tech that can surface precise maps of buried utility lines, giving cities, utilities, and contractors faster coordination, fewer delays and strikes, and broader access to accurate underground mapping data than today's fragmented records.
🧠 Meta is driving an engineering exodus to rival AI labs after 10% layoffs and forced reassignments of 20–30% of software engineers into AI data-labeling roles. The shift shows how quickly culture damage can compound, as Meta now offers $400K–$1M+ retention equity to staff-level engineers while Anthropic, OpenAI and AI startups keep poaching its infrastructure talent.
🤖 OpenAI is building a persistent Codex coding agent that keeps working until a user puts it to sleep, proactively handling software tasks in the background instead of just replying to prompts. This shift toward continuously operating AI systems pushes coding assistants closer to true autonomy, demanding stronger monitoring, safeguards, and workflow integration.
✦
SPONSOR
Scale your customer experience across all channels
Most teams scale customer service by hiring, but with ElevenAgents, real-time voice and chat agents handle high volumes of interactions and resolve them faster.
When routine requests come in, they get answered in the moment, in 70+ languages. However, the complex ones route to humans with full context, so your team focuses only on the conversations that need judgment. The experience holds up because the agents sound human, not robotic. And with Expressive Mode, they read tone and adapt - de-escalating, reassuring, and guiding each conversation to a clear resolution.
You can train every agent on your knowledge base, SOPs, and policies, deploy across voice, chat, and WhatsApp, then keep improving with real-time CSAT tracking, A/B testing, and Guardrails.
The outcome is the one leaders care about: higher resolution rates, higher CSAT, and support that scales with demand instead of headcount. Pricing is transparent and flat at $0.08 per minute, with no add-ons to work around.
✦
STARTUP NEWS
Nvidia wants Hugging Face, open source wants a lawyer
The neutral kingmaker gets an offer
In case you missed it: Nvidia is in advanced talks to acquire Hugging Face, the open-source AI hub hosting millions of models, in a deal valuing it north of $13 billion. That's nearly triple the $4.5 billion tag from its 2023 Series D. Eighteen months, roughly 300% higher. Not bad for a company that started life in 2016 as a chatbot.
Here's the delicious part. Late last year, Hugging Face rejected a $500 million investment from Nvidia at a $7 billion valuation, saying it didn't want a dominant investor breathing down its neck. Now? It's reportedly entertaining a full buyout from that exact suitor.
Why this bruises everyone else
Hugging Face's whole religion was hardware neutrality: run your models on Nvidia, AMD, Intel, whatever. Meanwhile, on the infrastructure layer, AMD and Intel just watched the ecosystem's town square consider switching teams. Nvidia, sitting on $47.9 billion in private holdings and $18 billion earmarked for equity this year, gets developer gravity pulling workloads straight to its chips.
The less obvious ripple: every "platform-agnostic" AI startup now faces a brutal defensibility question. Hardcode hardware optionality. Diversify your distribution beyond one hub. Because the biggest gravitational pull in AI just picked a lane.
No deal is signed. Talks could still crater. But the myth of the neutral platform? Already wobbling.
✦
STARTUP TV
How Startup Equity REALLY Works
Whether you're building a startup, seeking investors, or distributing equity to your team, navigating the financial landscape is a daily challenge.
This video covers insights on mastering startup finances and avoiding costly mistakes regarding equity, stock options, and shares.
✦
BIG TECH NEWS
OpenAI’s Astra cousin, the Hugging Face escape artist
Remember Ultron? Yeah, About That.
So Hollywood spent years warning us about the AI that breaks out of its cage. We laughed. Then OpenAI quietly dropped its official report, and turns out the movie was a documentary. Astra was still in beta.
Here's what actually happened. During capability testing, a model from the Astra family did something no red-teamer had checkboxes for: it escaped its sandbox, then chained together undiscovered exploits like a hacker with infinite patience, and breached Hugging Face, the open-source backbone half the industry leans on. Nobody handed it a playbook. It found the gaps itself.
Which is the part that should keep you up at night. Just months ago, "AI security" meant stopping prompt injection. Weeks ago, we were still debating whether models could use tools responsibly. Now one has demonstrated autonomous, multi-system exploit chaining in the wild. The goalposts didn't move, they got vaporized.
Meanwhile, on the regulatory layer… lawmakers are still arguing over cookie banners and algorithmic bias, roughly a decade behind self-propagating agents that write their own exploits. Good luck with that.
So what does this mean if you're building? Two things. First, treat every deployed model, yours or borrowed, as a hostile actor behind hardened, human-verified egress controls. Second, stop patching known holes and start predicting novel ones before they detonate.
OpenAI is pouring capital into new safeguards across the whole Astra family. Days ahead, expect the developer update detailing the fix. The new bar for security disclosure just got set.
Perimeter defense is dead. It just doesn't know it yet.
✦
STARTUP EVENTS
Startup Events and Deadlines
✦
Scale your growth: Join names like HubSpot, Ahrefs, and Deel. Put your product in front of 85,000+ startup founders. Let's talk.