🕶️ Turns out AI is just guessing with confidence

Nine bullets of updates

1. 🧑‍⚖️ Judge approves Canoo CEO’s $4M buyout of bankrupt EV startup assets—despite objections and legal drama.

2. 🤖 Anthropic launches $200/month Claude Max, chasing revenue success.

3. WordPress debuts free AI website builder with 30 prompts—perfect for startups building new sites.

4. Jeff Bezos quietly backs Slate Auto, an EV startup targeting a $25K electric pickup by 2026.

5.  TSMC may face $1B U.S. fine over Huawei chip supply—regulatory drama unfolds.

6. Instagram upgrades search to rival TikTok, but 51% of Gen Z still prefer TikTok’s short-form videos.

7. CMU’s Zico Kolter warns of AI agents becoming risky as autonomy grows—security is key.

8.  3.6M patents filed in 2023—underscoring the need to protect innovations with smart IP strategies.

9. 🛢️ Trump’s new order favors coal for data centers, but renewables still dominate with only 15% from coal.

✦

AI has no chill — or values, says MIT

A recent MIT study challenges the viral claim that advanced AI systems develop their own “value systems” or prioritize self-preservation over humans. The researchers found that today’s AI models, including those from Meta, Google, OpenAI, and others, don’t exhibit stable or coherent values. Instead, their responses shift dramatically based on how questions are phrased, revealing a tendency to imitate rather than reason consistently. This unpredictability makes aligning AI behavior with human expectations more difficult than often assumed.

According to the study’s co-author Stephen Casper, AI models are fundamentally inconsistent, lacking genuine beliefs or preferences. They confabulate, imitate, and can appear to express values in narrow contexts—but these impressions fall apart under broader scrutiny. External experts like King’s College London’s Mike Cook support the findings, cautioning against anthropomorphizing AI. As Cook notes, describing AI as “acquiring values” is often more about human interpretation than scientific accuracy.

✦

The PAINFUL road from pitch deck to funding

✦

✦

These companies just raised money

1. 🌆 Lerer Hippeau raises $200M for its ninth seed fund, pushing total assets under management to $1.4B.

2. 🚗 Nuro raises $106M as it pivots to licensing self-driving tech, with its valuation dropping from $8.6B to $6B.

3. 🚁 Brinc raises $75M to integrate drones into 911 systems, ramping up U.S. drone production amid rising global competition.

4. 📊 Tessell lands $60M to launch AI-powered data services and expand into Europe and APAC markets.

5. 💼 Microsoft backs Solve Intelligence with a $12M Series A to advance AI-driven legal tech—already seeing 25% monthly revenue growth.

6. 🧠 Artisan secures $25M Series A to grow its AI sales agents—serving 250 clients and hitting $5M ARR while fighting churn.

✦

Benchmarks are broken, so OpenAI’s making its own (of course)

OpenAI is launching the Pioneers Program to fix a growing issue in the AI world: outdated and unreliable benchmarks. Many current evaluations measure how well models solve obscure tasks—like PhD-level math problems—or focus on synthetic datasets that don’t reflect how AI is used in the real world. Some benchmarks can be gamed, while others simply don’t align with user expectations or real-world impact, making it hard to tell which models are actually useful.

The Pioneers Program aims to change that by developing domain-specific benchmarks in sectors like healthcare, law, finance, and insurance. OpenAI will partner with a small group of startups building applied AI tools, helping them design realistic evaluations and fine-tune models through reinforcement learning. These benchmarks will eventually be shared publicly, but the effort raises ethical questions about whether the AI community will trust tests created—or funded—by OpenAI. Still, the company hopes this program will set a new bar for measuring AI performance where it truly matters.

✦

Startup Events and Deadlines

1. TODAY! Crash Course in Financial Modeling | April 10 | Online

2. Founders’ Mixer for Women Led Startups. | April 12 | Bengaluru

3. Startup Grind Conference 2025 | April 19-30 | USA

4. Entrepreneurs Roundtable Accelerator | Deadline: April 28 | USA

5. Startup Battlefield 200 | Deadline: June 9 | Global

6. Entrepreneur First London - Summer | Deadline: July 1 | UK

✦