Morning Edition

The Futurist

Saturday, October 31, 2026
AI & Technology Markets & Crypto Ideas Worth Keeping
S&P 500$7,720.39▼ 0.36%
Dow Jones$52,342.56▼ 0.34%
Nasdaq$26,184.72▼ 1.42%
VIX$18.15▲ 5.46%
BITCOIN$77,611.00▼ 3.07%
RIPPLE$1.26▼ 2.78%
ETHEREUM$2,222.31▼ 4.11%
DOGECOIN$0.08▼ 3.55%
01

AI & Technology

COINDESK

Binance's $6.3 Billion Hack Just Got Worse

Binance has uncovered a broader breach than initially disclosed in what is now the largest crypto hack in history — the stolen Bitcoin alone exceeds Tesla's entire holdings and is more than double El Salvador's reserve. Withdrawals were suspended as the exchange targets a Sunday restart, while Wall Street analysts are paradoxically urging investors to buy the dip on crypto stocks.

404 MEDIA

Proton CEO Cancels AI Re-Interviews After Public Backlash

Following a 404 Media investigation revealing that Proton CEO Andy Yen was forcing employees through AI-administered re-interviews amid internal fears about job security, Yen has reversed course and cancelled the program. A rare case of journalism directly killing a bad corporate policy in real time.

404 MEDIA

Apple Is Entering Its Slop Era

Apple, long the standard-bearer of taste and restraint, is now embracing the AI content slop cycle that has degraded nearly every other platform. The piece argues this represents a meaningful shift in Apple's identity — not a bump, but a direction.

SIMON WILLISON

Someone Baked a Working NES Emulator Into an LLM's Weights

Drew Breunig embedded a functional 6502 CPU emulator — the chip that powered the Nintendo Entertainment System — directly into the weights of Qwen3.5-4B. You can now run Super Mario Bros inside a language model, slowly. This is less a product and more a proof of concept that should make everyone rethink what "model weights" actually means.

COINDESK

OpenAI Cuts Fundraising Target by $10B, Valuation by $150B

OpenAI has slashed its fundraising ambitions significantly amid a broader tech selloff and what markets are calling "AI jitters." This lands alongside the Anthropic IPO buzz — a stark contrast in investor sentiment between the two leading AI labs right now.

A16Z

Why Frontier Models Still Can't Parse a PDF

Reducto CEO Adit Abraham makes the case that the document layer — PDFs, scans, spreadsheets — remains the unsolved bottleneck keeping AI agents out of healthcare, legal, and finance. The reasoning stacks work great; feeding them real-world data still doesn't.

02

X / Twitter Signal

OpenAI has cut its funding target by $10 billion and its valuation by $150 billion in a single week, per the Financial Times. This isn't a rounding error — it's a signal that the market's patience with AI infrastructure burn rates is compressing fast, especially as the Nasdaq drops 1.42% and VIX climbs above 18.

a16z flagged what they're calling the "Rack Spread": B300 GPU rental prices up 106% since mid-July, while token prices fall. Translation — the cost of running AI is rising while the price you can charge for AI output is dropping. That's a margin squeeze hiding inside the capex boom, and it hasn't shown up in most investor models yet.

Ethan Mollick catalogued OpenAI's product graveyard: Sora lasted 6 months, Agent Builder 8 months, Group Chat 8 months, Atlas 9 months. The pace of kill-and-ship is accelerating, and Mollick's implication is sharp — most of these weren't failures of execution, they were failures of market fit that nobody waited long enough to discover.

Demis Hassabis retweeted Mercor's announcement that Gemini 4 Carbon debuted at #1 on the APEX-Agents leaderboard — the only model with a mean score above 90%. Quietly, Google DeepMind is winning the agentic benchmark race while the OpenAI valuation drama dominates headlines.

03

The Thread

Two numbers define this week: $6.3 billion stolen from Binance, and $150 billion wiped from OpenAI's valuation. Neither happened because the underlying technology broke — they happened because trust did. Binance's breach is metastasizing (the hack keeps getting bigger with each disclosure), and OpenAI is discovering that "the most valuable startup in history" is a title markets can revoke quickly when sentiment turns.

The cost of compute is rising. The price of AI output is falling. Someone is going to be left holding that gap.
The a16z "Rack Spread" data is the most important chart nobody is talking about. GPU rental prices up over 100% while token prices drop means the AI infrastructure trade is quietly inverting. The labs and hyperscalers assumed they could charge more as models got better — but commoditization is arriving faster than monetization. OpenAI's fundraising haircut isn't just market noise; it's the first public acknowledgment that the math is harder than the pitch decks suggested. Meanwhile, the NES-in-a-language-model story is a useful corrective to all the doom. Someone just ran Super Mario Bros inside model weights as a weekend project. The distance between "this is impossible" and "this is a GitHub repo" keeps shrinking. Whatever the funding environment does, the capability curve isn't waiting for Wall Street to sort out its feelings.