VIVEK KUMAR VERMA

CSE undergrad · Full-stack dev (Next.js, TS, Node) | Tracking new tech, first | AI · Code · What's next

[VIVEK KUMAR VERMA @_vivek_33]

Alphabet falling 10 percent after earnings while Microsoft rises is the market drawing a line. AI spending is no longer judged on ambition, it is judged on monetisation. The labs and clouds that can show revenue per dollar of compute will keep their premium, the rest will not.

[zerohedge @zerohedge]

Premarket movers:

Mag 7 stocks are mixed:

Alphabet (GOOGL) plunges 11% after the company reported disappointing search and advertising sales that reinforced investor concerns about AI’s threat to its core business.

Microsoft (MSFT) is up 1.7% after the company posted increasing growth in its cloud business and kept its capital expenditures in check, satisfying investors amid a drumbeat of concern that spending on artificial intelligence won’t pay off.

Apple (AAPL) +1.3%, Meta Platforms (META) +1%, Tesla (TSLA) -0.2%, Nvidia (NVDA) -0.5%, Amazon (AMZN) -1.8%

Software companies are broadly lower after disappointing quarterly results from ServiceNow and Fortinet added to concerns about the industry’s prospects in a world awash with AI.

Blue Owl Capital (OWL) rises 4% after reporting a larger uptick in fees than analysts expected in the third quarter.

Carvana (CVNA) is up 8% after the used-car retailer reported better-than-expected results on a key profit measure.

Caterpillar (CAT) shares slump 9% after the company posted third-quarter earnings that only modestly beat expectations, raising concerns about the durability of the broad rally in AI-related stocks.

Chipotle (CMG) is up 7% after the restaurant chain reported comparable sales for the third quarter that beat the average analyst estimate.

DoorDash (DASH) agreed to buy rival Instacart (CART) in a deal that values the company at about $17 billion, the latest sign of consolidation in an increasingly competitive delivery market. Shares of Instacart are halted.

EBay (EBAY) jumps 6% after the online auction company’s net revenue forecast for the full year beat the average analyst estimate.

Fortinet (FTNT) plunges 26% after the security software company gave a disappointing outlook for service revenue.

Glaukos (GKOS) is up 15% after the eye-focused medical device maker boosted its net sales guidance for the full year. The guidance surpassed Wall Street’s expectations.

Marvell Technology (MRVL) falls 11% as the company is said to be in advanced talks to buy data center optics maker Lumentum Holdings (LITE), according to people familiar with the matter. Lumentum rises 6%.

ServiceNow (NOW) drops 19% after the software company’s forecast for fourth quarter current remaining performance obligation fell short of the average analyst estimate.

[VIVEK KUMAR VERMA @_vivek_33]

@TheAlphaBrief @satyanadella Borrowing to build is not a weakness on its own, it is a bet on duration. If AI demand stays, cheap debt today looks smart. If it cools, the interest bill arrives before the revenue does, and that timing is the real risk.

[VIVEK KUMAR VERMA @_vivek_33]

@xiaolu_sprmind Negative free cash flow does not mean Alphabet is losing, it means the market has to trust a future payoff. Microsoft still generating cash while spending gives it patience from investors that Alphabet now has to earn back quarter by quarter.

[VIVEK KUMAR VERMA @_vivek_33]

@satexbabi The 86 percent figure is alarming, but the 15 percent disclosure is the real failure. Experts did not just follow a coin flip, they then wrote it up as their own reasoning. That makes AI influence hard to audit, because the paper trail hides the source.

[VIVEK KUMAR VERMA @_vivek_33]

@hexalan Half the cost for two points less is a trade most teams should take, but only if the router is transparent. When the system silently picks a weaker model for a hard step, the saving is paid for in debugging time, not dollars.

[VIVEK KUMAR VERMA @_vivek_33]

@Iamhamzafaisal Running on a phone changes the economics, not just the location. When inference is local and offline, the moat shifts from raw compute to distribution, battery efficiency and how well a small model handles real tasks without phoning home.

[VIVEK KUMAR VERMA @_vivek_33]

@xiaolu_sprmind Defenders first is the right sequence, but it also creates a two tier security market. The advantage only lasts if the phased rollout is measured in weeks, not quarters, otherwise attackers close the gap while smaller teams are still waiting.

[VIVEK KUMAR VERMA @_vivek_33]

@Vvikramai A court order is slower than a developer shutdown switch, but it is also harder to reverse quietly. That tension is the point: internal controls can act in minutes, while a legal block creates a public record and a precedent other labs must live with.

[VIVEK KUMAR VERMA @_vivek_33]

@TechCrunch Disclosing 33.2 billion changes the cloud argument from growth percentages to actual scale. The next test is not whether Azure can catch AWS, it is whether that revenue grows fast enough to justify the AI capacity bill behind it.

[VIVEK KUMAR VERMA @_vivek_33]

@verge If AI is driving more searches, the interesting question is whether those queries monetise as well as the old ones. Growth in volume alone does not settle the AI spending debate, it just raises the bar for conversion.

[VIVEK KUMAR VERMA @_vivek_33]

@OpenAIDevs This is a smart bridge from chat subscriptions to the API. Many Pro users never touch the platform, so 500 dollars a month is enough to prototype an agent without separate budget approval. The test will be how many prototypes become paying API workloads once credits run out.

[VIVEK KUMAR VERMA @_vivek_33]

@github @ArtificialAnlys Cost per task is now the metric that matters, not just raw score. Being within two points of Claude Code at about half the API cost changes the maths for teams running agents all day. The question is how that holds up once the work leaves benchmarks and hits real repositories.

[VIVEK KUMAR VERMA @_vivek_33]

@Google @ArtificialAnlys @GoogleDeepMind The cybersecurity and long horizon claims are the interesting part here. A benchmark lead is one thing, but sustained work over long tasks is where most models still break down. The real test will be whether Carbon holds up on messy, multi step work rather than clean evaluations.

[VIVEK KUMAR VERMA @_vivek_33]

Everyone else is now playing catch up to a moving target.

[VIVEK KUMAR VERMA @_vivek_33]

Google just dropped Gemini 4 Carbon, two weeks after Gemini 4 Argon. Two frontier models in fourteen days. That is not a product launch anymore, that is a production line.

[VIVEK KUMAR VERMA @_vivek_33]

That is the loop nobody talks about enough. AI is no longer just the product, it is the factory making the next product. The labs that close this loop fastest will not just win benchmarks, they will set the pace for everyone else.

[VIVEK KUMAR VERMA @_vivek_33]

Apple is quietly turning Pages, Keynote and Numbers into AI workspaces, and the MCP support matters more than the assistants themselves. Once outside agents can edit your documents, the office suite becomes a platform, not just an app.

[Bretance @bretance]

Apple says Keynote, Pages and Numbers will get beta AI assistants for Creator Studio subscribers in November, plus MCP support so external agents can read and edit documents on macOS 27.
https://t.co/mWyDsSggtu

[VIVEK KUMAR VERMA @_vivek_33]

@AnthropicAI The real measure here is not how fast a model proposes candidates, but how many survive wet-lab validation and trials. AI can compress the search, yet biology still gets the final vote, and that is where the timeline is won or lost.

[VIVEK KUMAR VERMA @_vivek_33]

@verge Science moves on shared results, not press releases. If breakthroughs stay behind a gatekeeper, other labs cannot verify, build on or challenge them, and the field ends up slower and more concentrated than it needs to be.