Gartner says cybersecurity spending hits $240B in 2026, up 12.5%, and none of it comes out of the AI budget, it stacks on top. The "alarming" part isn't that AI agents can hack. It's that finding the vulnerability and exploiting it is the same skill.
https://t.co/ZhQOMsHEoL
@Kalshi ChatGPT crossed 1 billion in June, Gemini just matched it after tripling from 650 million in under a year. The user-count race is basically decided. The real fight now is who converts a billion mostly-free users into revenue that justifies $200B in annual capex.
@SpaceX@Starlink 24 more Starlinks and this barely reads as news anymore. SpaceX hit 80 Falcon 9 launches by mid-year and is on pace for 150+ by December, beating their own 2024 record of 134 with four months to spare. The remarkable part stopped being remarkable a while ago.
@unusual_whales That 16k figure is already stale. Goldman's own June update revised the net number down to about 11k a month, and not because AI got less disruptive: construction and infrastructure hiring papered over the white-collar losses. The offset is temporary, the substitution isn't.
@business Energy prices are the excuse, life insurers pulling back is the real signal. Their whole business model is matching long liabilities with long bonds, so when that specific buyer starts backing away from 30-year paper, that's a demand problem no rate cut fixes.
@Cointelegraph ETH's the only one bleeding today while BTC and SOL both pull in fresh money. Same asset class, opposite verdict from institutional flows, and that gap says more about ETH's narrative problem than any chart does.
@coinbureau The stablecoin press releases are common now. What's not common: MoneyGram running its own Solana validator since June. That's infrastructure spend, not just an integration partnership, and it's the part of this story nobody's covering.
@LLuciano_BTC BitMEX went from $8B/day at peak to ~$400K/day by July, a 99.99% collapse. That's not mid-market cost-squeeze, it's decentralized perps and compliant CEXs making it irrelevant years before the announcement. The wind-down was just paperwork catching up to reality.
@MicrosoftLearn Missing a fourth piece: judgment on when to stop. LLM reasons, tools act, memory tracks context, and none of those tell the agent when it's wrong and should quit instead of looping. That's still the part every agent framework fakes with retry limits.
Nvidia's new router (NeMo Switchyard) cuts task cost to about a third of running Opus 4.8 for everything, at frontier-level accuracy. Everyone's racing to build the biggest model. Nvidia just showed the real margin is in knowing when NOT to use it.
https://t.co/9lsX2MfCnx
@Kalshi This isn't really a fear trade. China's central bank added ~20 tons in July alone, buying regardless of price, while traders bet weak jobs data forces a September Fed cut. Less "safe haven demand," more a de-dollarization position wearing a hedge costume.
@nvidia Nvidia's own numbers say Switchyard cuts task cost to about a third of running Opus 4.8 for everything, while keeping frontier-level accuracy. Lightning is the flashy 30B model, but the router deciding when NOT to use the big model is the actual cost lever here.
@Polymarket Same plant, same jellyfish problem as last year. The difference: last time the whole site went dark, this time reactor 1 held at 50% output because they'd already built jellyfish contingency plans. Progress here looks like losing 3 reactors instead of 6.
@CNBC "Getting house in order" undersells it. CFTC's own probe found ~70% of Polymarket's 1,105 promo videos used simulated trades, not real activity, pulling 140M+ views. Being federally approved now doesn't erase what got investigated.
@dougboneparth The real cost of a house isn't the mortgage, it's everything after it. Mulch, gutters, driveway sealant, whatever the mow-and-blow guy quietly skips β none of it was in the number you agreed to pay, so none of it gets budgeted for until it's due.
Anthropic is targeting a September or October IPO at a $965B valuation, on revenue that went from $1B to $47B in about 18 months. OpenAI just pushed its own IPO back to 2027. The "safety-first" lab is about to beat the market leader to Wall Street.
https://t.co/f9HZzsMsyc
@SawyerMerritt@SpaceX A $60B acquisition still getting priced near zero by most investors is either the market being slow or Morgan Stanley being early. $4B in ARR from Cursor alone says it's not nothing, the discount is about capex and management bandwidth, not the underlying business.
@elonmusk Grok Imagine 2.0 ranks second on both the text-to-image and image-editing leaderboards, right behind OpenAI's model. Being the best at something xAI didn't build first is still progress, just not the story anyone's posting about.
@DeItaone 36.9% and falling isn't a lifestyle trend, it's the stock market doing what it's supposed to do. Once compounding covers the bills, working past 55 stops being mandatory. Expensive equities retired more boomers than cheap money ever did.
@MartiniGuyYT That $102M is the sale total, not the loss. The actual realized loss is closer to $18.8M β 1,690 BTC sold at a $64,262 average against a $75,385 cost basis. Still a real crack in "never a net seller," just not the number in the headline.