Sunday Jul 26
RUMORSOther

Stripe In Talks To Buy OpenRouter

26JUL
$10B, UNSIGNEDSTRIPEOPENROUTER

Stripe wants to own the AI plumbing, not just the payments. The Wall Street Journal reports Stripe is nearing a $10B buy of OpenRouter. Nothing is signed. Rivals could still jump in and outbid Stripe.

OpenRouter lets developers switch between hundreds of AI models through one API.

Stripe already processes OpenRouter's payments; this would make it the owner.

Co-founder Alex Atallah calls OpenRouter the Stripe of the AI field.

The price would be triple OpenRouter's $1.3 billion valuation from May.

Big Tech rivals have also circled the same asset.

Talks are fluid and could still fall apart before signing.

If it closes, payments and AI model routing become one company's business.

full brief & sources

⚡ Why this matters

  • Payments companies are racing to own the AI infrastructure layer, not just process cards.
  • A confirmed deal would triple OpenRouter's valuation in under three months.
  • It signals AI model routing is now a strategic asset, not just plumbing.

🔍 What happened

  • The Wall Street Journal first reported the talks on July 24, 2026.
  • OpenRouter lets businesses compare and switch between hundreds of AI models.
  • The reported price is close to $10 billion, up from a $1.3 billion valuation in May.
  • Stripe already relies on OpenRouter as a customer for payment processing.
  • No final agreement has been signed, and other bidders have reportedly circled OpenRouter.

💬 Smart takes

  • Alex Atallah, OpenRouter co-founder: calls the platform the Stripe of the AI field, a single entry point so businesses don't get locked into one model.
  • Patrick Collison, Stripe CEO: has said AI agents will handle most online transactions, and Stripe has been building infrastructure to match.
  • Skeptic: a $10 billion price for a company valued at $1.3 billion two months ago smells more like FOMO than fundamentals.

🧭 Where this goes

  1. Likelythe deal, if it closes, folds OpenRouter into Stripe's existing developer platform.
  2. Possiblea rival bidder, a hyperscaler or another payments firm, tops Stripe's offer.
  3. Possiblethe valuation jump becomes the new comp for AI-routing infrastructure startups.
  4. Wild Cardtalks collapse entirely and OpenRouter raises another round at a lower price instead.

🥄 The Spoon Take

Stripe running your AI model calls the same way it runs your card swipes matters more than the price. If this closes, the company behind half the internet's checkout also decides which AI model your app uses.

🤔 Pushback

This is still just reported talks, not a signed deal, and rivals could outbid Stripe. A $10B price for a $1.3B company also invites real diligence risk.

·Axios - Stripe's AI currency move·Yahoo Finance - Stripe eyes OpenRouter·PYMNTS - Stripe doubles down on AI
Tuesday Jul 14
RUMORSGoogle

Gemini 3.5 Pro Leaks Ahead Of Launch

14JUL
LEAKEDTOO SOON

Google's next flagship model is leaking before it's confirmed. Leakers say Google scrapped the original base and rebuilt it from scratch. Unconfirmed specs: a 2-million-token window, target date July 17.

Google has not confirmed a single detail. Every spec below comes from unnamed sources and leaked screenshots, not an official post.

DeepMind reportedly scrapped the first 2.5 Pro base after finding bugs in tool-calling and SVG generation. The rebuild runs on a native Gemini 3 foundation. Deep Think reasoning stays locked to the $250 Ultra tier.

July 17 is the target, July 24 the fallback. DeepSeek's own release deadline lands right in between.

full brief & sources

⚡ Why this matters

  • Google keeps missing its own frontier-model timeline while rivals ship weekly.
  • A scrapped base model suggests Gemini 3 had real technical problems.
  • The leak sets a specific date the market will now hold Google to.

🔍 What happened

  • Leaks point to a July 17 general-availability date for Gemini 3.5 Pro.
  • Google reportedly discarded its original 2.5 Pro base after tool-calling and SVG bugs.
  • The rebuild uses a native Gemini 3 foundation, per internal sources.
  • Unconfirmed specs list a 2-million-token context window, the largest of any frontier model.
  • Deep Think reasoning is rumored to gate behind the $250-per-month Ultra tier.
  • No model card, pricing page, or API listing exists publicly as of July 13.

💬 Smart takes

  • Tech Times: every specific claim about the date, context window, and pricing comes from third-party reporting, not Google.
  • X leaker Pankaj Kumar: frontend generation sees a major jump, with stronger SVG output and one-shot game demos.
  • Skeptic: Google has slipped this exact launch before; a scrapped base model could mean more delay, not a stronger product.

🧭 Where this goes

  1. LikelyGoogle either confirms or quietly misses the July 17 date within days.
  2. Possiblethe 2-million-token context window ships but with degraded quality at full length.
  3. PossibleDeep Think stays Ultra-exclusive to protect margin on the $250 tier.
  4. Wild CardGoogle preempts the leak with an early announcement to control the narrative.

🥄 The Spoon Take

A leak this detailed, four days before a launch nobody confirmed, is its own signal. Google is now competing against its own rumor. If Gemini 3.5 Pro under-delivers on July 17, the leak becomes the story instead of the model.

🤔 Pushback

Leaked specs from unnamed sources have been wrong before, and 'scrapped the base model' is exactly the kind of dramatic detail that gets exaggerated in the retelling.

·Tech Times - Gemini 3.5 Pro leak·Geeky Gadgets - Deep Think leak details·X - Pankaj Kumar leak thread
Saturday Jul 11
RUMORSOther

DeepSeek Is Reportedly Building Its Own Chip

11JUL
OWN CHIP?DEEPSEEK

China's most-watched AI lab wants off Nvidia's leash. DeepSeek is reportedly designing its own inference chip, Reuters says. Export controls already block Nvidia's best chips from China.

The report says the lab has spent about a year lining up design, foundry, and supply partners. The silicon is said to target the model-serving stage, not training new systems.

US trade restrictions already keep the industry's top training hardware out of the country. Until now the lab has depended on two foreign suppliers for its models. OpenAI already shipped its own model-serving silicon with Broadcom last month.

Neither side has confirmed any of this. If accurate, nearly every frontier lab is now quietly designing its own silicon instead of just buying it.

full brief & sources

⚡ Why this matters

  • Export controls make Nvidia's best chips off-limits in China, so a homegrown chip would remove DeepSeek's biggest supply risk.
  • If confirmed, DeepSeek joins OpenAI and reportedly Anthropic in the in-house-silicon trend, right as GPU supply stays tight everywhere.

🔍 What happened

  • Reuters reported the chip news on July 7, citing three unnamed sources.
  • The chip reportedly targets inference workloads, not model training.
  • DeepSeek has been talking to chip-design, foundry, and memory partners for about a year.
  • US export rules already block Nvidia's most advanced training and inference chips from Chinese buyers.
  • OpenAI unveiled its own inference chip, Jalapeño, with Broadcom on June 24.
  • Reuters separately reported in April that Anthropic is weighing building its own chips.

💬 Smart takes

  • Reuters: DeepSeek has been in talks with chip-design, foundry, and memory partners for about a year.
  • Skeptic: designing a chip and shipping a competitive one are very different things, and neither DeepSeek nor Nvidia has confirmed a word of this.

🧭 Where this goes

  1. LikelyDeepSeek neither confirms nor denies the report for now, standard practice for in-progress silicon.
  2. Possiblea confirmed DeepSeek chip surfaces within 12 months, following OpenAI's Jalapeño timeline.
  3. PossibleHuawei ends up supplying the fabrication DeepSeek can't get from Nvidia-aligned foundries.
  4. Wild CardDeepSeek's chip underperforms and the story quietly disappears, like several rumored Chinese chip projects before it.

🥄 The Spoon Take

Every major AI lab is quietly becoming a chip company. OpenAI already shipped one, Anthropic is reportedly weighing it, and now DeepSeek. Export controls didn't stop China's favorite model, they just pushed it to build the one thing Nvidia can't sell it.

🤔 Pushback

Reuters cites three anonymous sources, and neither DeepSeek nor Nvidia confirmed anything. 'In talks for a year' is a long way from a shipped chip.

·Bloomberg - Reuters own-chip report·US News - Exclusive sources say·TechCrunch - OpenAI's Jalapeño chip
Wednesday Jul 8
RUMORSMeta

Meta Claims Watermelon Matches GPT-5.5

2JUL
METAGPT-5.5

Meta says its next model finally caught up to OpenAI's best. Alexandr Wang, Meta's AI chief, reportedly told staff Watermelon matches GPT-5.5. No named benchmarks, no outside check yet.

Watermelon is the successor to Avocado, the model behind Muse Spark.

It reportedly uses 10 times more compute than its predecessor.

Wang made the claim at an internal town hall, not a launch.

He didn't name which benchmarks Watermelon supposedly matches.

Meta also teased a faster Muse Spark coding update coming soon.

This is the fourth Meta AI claim this year without independent verification.

If true, it ends OpenAI's run as the undisputed benchmark leader.

full brief & sources

⚡ Why this matters

  • An unverified benchmark claim from a lab chief can move competitive narratives overnight.
  • Meta has a track record of teasing capability before shipping it.
  • If real, it changes who enterprises trust for their next model migration.

🔍 What happened

  • Alexandr Wang told an internal Meta town hall on July 2 that Watermelon matches GPT-5.5.
  • Watermelon is Meta's next flagship model, successor to Avocado (Muse Spark's codename).
  • Wang said Watermelon uses roughly 10x the compute of its predecessor.
  • No specific benchmark names or scores were disclosed.
  • Business Insider sourced the claim from people familiar with the town hall.
  • OpenAI has not commented on the claim.

💬 Smart takes

  • Business Insider: the claim came from 'one executive, in one internal room, on one round of unnamed benchmarks.'
  • Alexandr Wang: a Muse Spark coding and agent update is coming 'pretty soon.'
  • Skeptic: Meta has claimed benchmark parity before and shipped models that fell short in independent testing.

🧭 Where this goes

  1. LikelyMeta publishes at least partial Watermelon benchmarks within the next 60 days.
  2. Possibleindependent evaluators test Watermelon and find a smaller gap than claimed.
  3. PossibleOpenAI responds with its own updated GPT-5.5 benchmark refresh.
  4. Wild CardWatermelon ships and actually beats GPT-5.5 on reasoning, not just matches it.

🥄 The Spoon Take

An internal town hall claim isn't a launch, but it's a signal Meta wants told. Every 'we caught up' leak before an actual release buys goodwill cheaply. The real test is the benchmark table Meta eventually has to publish.

🤔 Pushback

This is a single executive's claim in a private room, with no named benchmarks and no outside verification. Meta has made similar catch-up claims before that didn't hold up once independent labs ran the numbers.

·Business Insider via Benzinga - Wang's claim·Implicator - Wang vs Zuckerberg on agents·AI Weekly - Wang's Watermelon alert