> ## Content Index
> Fetch the complete content index at: https://www.implicator.ai/llms.txt
> Use this file to discover other available public pages before exploring further.

# Ramp raises $1.85B; Meta's AI trails three Chinese models on its own chart
- URL: https://www.implicator.ai/ramp-1-85b-meta-open-model-trails-chinese-models/
- Published: 2026-10-16T11:45:00.000Z
- Updated: 2026-10-16T11:45:00.000Z
- Description: Modal says its sandboxes for AI-written code are a $300 million annualized run-rate product on their own.
- Author: Marcus Schuler
- Tags: Morning Briefing

**IMPLICATOR** **​.ai**

Morning Briefing · From San Francisco

Friday, October 16, 2026

9 stops

From San Francisco

| 1 | The Editorial |
| - | ------------- |

*Morning, humans.*

*Each of today's three picks comes with a condition in the small print.*

*Ramp announced a $1.85 billion round on Oct. 15, two days after opening its internal AI workspace to outsiders. The finance company has an AI bill worth examining.*

*Meta released a model with 816 billion parameters on Oct. 15\. Its weights are available to download, though anyone planning to sell access should read the terms first.*

*Amazon's $74.99 Bee Frontier comes with 14 days of memory. Panos Panay's talk of a future beyond apps invites a closer look at where Bee puts its summaries.*

*Stay curious,*

*Marcus Schuler*

Good briefing? Pass it on

[X](https://twitter.com/intent/tweet?text=Strategic%20AI%20news%20from%20San%20Francisco.%20No%20hype%2C%20just%20what%20moved%2C%20who%20won%2C%20and%20why%20it%20matters.&url=https%3A%2F%2Fwww.implicator.ai%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dshare%26utm%5Fcampaign%3Dreader%5Fshare%26utm%5Fcontent%3Dtwitter&ref=implicator.ai) · [LinkedIn](https://www.linkedin.com/sharing/share-offsite/?url=https%3A%2F%2Fwww.implicator.ai%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dshare%26utm%5Fcampaign%3Dreader%5Fshare%26utm%5Fcontent%3Dlinkedin&ref=implicator.ai) · [Bluesky](https://bsky.app/intent/compose?text=Strategic%20AI%20news%20from%20San%20Francisco.%20No%20hype%2C%20just%20what%20moved%2C%20who%20won%2C%20and%20why%20it%20matters.%20https%3A%2F%2Fwww.implicator.ai%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dshare%26utm%5Fcampaign%3Dreader%5Fshare%26utm%5Fcontent%3Dbluesky&ref=implicator.ai) · [Email](mailto:?subject=A%20newsletter%20I%20think%20you%27d%20like&body=This%20is%20one%20of%20the%20few%20AI%20newsletters%20I%20actually%20read.%20Strategic%2C%20no%20hype.%0A%0ASubscribe%20free%3A%20https%3A%2F%2Fwww.implicator.ai%2Fsubscribe%2F%3Futm%5Fsource%3Dnewsletter%26utm%5Fmedium%3Dshare%26utm%5Fcampaign%3Dreader%5Fshare%26utm%5Fcontent%3Demail)

| 2 | THE BIG STORY |
| - | ------------- |

Ramp reaches a $60 billion valuation with AI spending controls on sale.

**Ramp announced a $1.85 billion round led by Dragoneer Investment Group on Oct. 15 at a $60 billion valuation.**

Ramp says annualized revenue topped $2 billion as of Oct. 1 and cash flow is positive.

Glass, Ramp's AI workspace, entered early access for outside companies Oct. 13 with per-project spending meters. Co-CEO Eric Glyman said Ramp spent $1.5 million on AI tokens in one week in May; his claim that Router cut Ramp's AI bill 30% has no dollar baseline.

**Why This Matters:**

- Finance teams paying AI bills can get per-project meters and spending limits in Glass's early access, with outside customer results still unverified.
- If Ramp's AI Token Spend Management catches on, finance teams could make token spending a routine budget item.

Reality Check

**What's confirmed:** Ramp announced $1.85 billion in primary financing on Oct. 15\. Glass is in early access.

**What's implied (not proven):** Ramp's internal experience may help other companies control AI costs.

**What could go wrong:** Glyman's 30% Router figure has no dollar baseline, and the 17.5% cut comes from an unnamed customer in Ramp's release. Outside companies may see smaller reductions.

**What to watch next:** Outside customer savings against a stated dollar baseline, with the amount of work held steady.

[Read the full story →](https://www.implicator.ai/ramp-glass-ai-workspace-token-spending/)

| 3 | ALSO TODAY |
| - | ---------- |

Meta ranks its “most capable US open-weight model” below three Chinese open models.

**Meta released Muse Spark Open's weights on Oct. 15, with 816 billion total parameters and 42 billion active per token.**

On the Artificial Analysis Intelligence Index in Meta's own table, the model scores 39, below 45, 44 and 40 for three Chinese open models. API sellers with more than $20 million in annual gross revenue must notify Meta and sign a separate agreement first. Buyers choosing by score have higher-scoring Chinese open options.

[Read the model card →](https://impli.me/ltGMhX?ref=implicator.ai)

| 4 | The Outside Read |
| - | ---------------- |

**Simon Willison's Weblog describes how he built a newsletter archive for his blog, starting with about half an hour of voice prompts while cooking dinner.**

He spent about another half hour typing prompts and reviewing, replacing a Git-subprocess import with an API-based version because one source was a private repository. “I usually cook with a podcast or TikTok running; now I can actually build stuff instead.”

[Read it at Simon Willison's Weblog →](https://simonwillison.net/2026/Oct/9/built-using-my-voice/?ref=implicator.ai)

| 5 | The One Number |
| - | -------------- |

$300M

Annualized run rate Modal reports for Sandboxes alone, a product launched in 2023 after it saw people using language models to write code. Modal reports close to 3 billion sandboxes, a company run rate above $1 billion and a $750 million Series D led by Accel. Executing generated code offers its own revenue stream.

Source: [Modal, Oct. 14, 2026](https://impli.me/DbrqC3?ref=implicator.ai)

| 6 | Today's Headlines |
| - | ----------------- |

- **Apple Intelligence** has [approval in China](https://www.implicator.ai/tim-cook-says-apple-intelligence-is-approved-in-china-without-siri-ai/), Tim Cook said Oct. 15, with a test version due later in October without Siri AI.
- **Zenzap** emerged from stealth Oct. 15 with [$44 million raised since 2022](https://www.implicator.ai/zenzap-44-million-ai-agents-frontline-chat/), targeting shift workers with a work messenger and AI agents.
- **Iambic Therapeutics** [priced its IPO](https://impli.me/lV8OgX?ref=implicator.ai) at $16 a share for $150 million in gross proceeds Oct. 14; Nasdaq trading began Oct. 15 under IAM.
- **YouTube Music** began [labeling tracks made primarily with generative AI](https://impli.me/nKAXxQ?ref=implicator.ai) on Oct. 14, with expansion to the main YouTube app planned.
- **Google** began rolling out [Universal Cart](https://impli.me/oCG15e?ref=implicator.ai) across U.S. Search and the Gemini app Oct. 15, combining products from multiple retailers while merchants remain sellers of record.

**From PRO:** In our Oct. 9 Mac Studio test, editing a 7,821-token prompt's last token cut Ollama's prompt processing from 8.51 seconds to 0.07; editing its first left 8.50 seconds. [See the server test →](https://www.implicator.ai/ollama-vllm-sglang-best-open-model-server/?utm%5Fsource=implicator&utm%5Fmedium=email&utm%5Fcampaign=pro%5Ffinding&utm%5Fcontent=ollama-vllm-sglang-best-open-model-server) PRO: $10 a month, $99 a year.

| 7 | The 5-Minute Skill |
| - | ------------------ |

**Build a travel recovery plan before the delay.**

A tight itinerary can fail after one late flight. Have the model find the breakpoints while changes are cheap.

**Your raw input:**

Gather your itinerary with dates, local times, reservation rules, transfer times, meeting deadlines, refundable options, and the latest acceptable arrival time for each commitment.

**The prompt:**

Review the itinerary and constraints I provide as a travel risk planner. Find every connection or reservation that could fail if an earlier segment is delayed by 30, 90, or 180 minutes. For each breakpoint, list the consequence, the decision deadline, and the lowest-cost recovery action available from the information provided. Then create a time-first fallback plan and a separate budget-first plan. Separate confirmed facts from assumptions. Do not invent schedules, prices, policies, or availability. End with a short checklist of details I should verify with the carrier, hotel, or meeting host before departure.

**Why this works:**

Delay scenarios make the model trace dependencies across the whole trip. The two fallback goals prevent one vague recommendation and expose where speed costs more.

**What to use:**

A reasoning model for the dependency check; a browsing model only for live schedules and prices, confirmed with the provider before booking.

| 8 | What To Watch Next |
| - | ------------------ |

| MON 10/19 | Fairs: Gartner opens IT Symposium/Xpo in Orlando, with its opening keynote at 10 a.m. ET and sessions running through Thursday.      |
| --------- | ------------------------------------------------------------------------------------------------------------------------------------ |
| MON 10/19 | Politics: The European Parliament begins its Strasbourg plenary session, which runs through Thursday.                                |
| TUE 10/20 | Finance: The Bureau of Labor Statistics reports September state employment and unemployment at 10 a.m. ET.                           |
| TUE 10/20 | AI: NVIDIA opens GTC Berlin with training workshops, followed by its Wednesday keynote and the conference and expo through Thursday. |
| FRI 10/23 | Tech: Apple makes the iPhone Duo, its first foldable iPhone, available.                                                              |

| 9 | The Rausschmeisser\* |
| - | -------------------- |

Amazon's $11.99 monthly Bee Pro tier comes with yet another app.

*Amazon introduced the $74.99 Bee Frontier on Oct. 15 with* [*14 days of memory included*](https://impli.me/3gwOpY?ref=implicator.ai)*. The device turns conversations into summaries, to-dos and reminders.*

**Our take:** Amazon's $74.99 Bee Frontier includes 14 days of memory, with summaries read in the Bee mobile app. On Oct. 15, devices chief Panos Panay imagined users who wouldn't ask for an app: “I think that is the next generation.” Bee gives that generation somewhere to read about what it just said.

For $11.99 a month, Bee Pro includes “the Bee desktop application” among its extra features. The generation that won't ask for an app can pay monthly for a second one. The future beyond apps appears to have an app department.

\*German for the last song of the night, the one that clears the room.

FREE WEEKDAY MORNING BRIEFING

Don’t miss the next AI story that matters.

The Implicator Morning Briefing filters the AI news cycle to the stories worth your attention and explains their consequences. From San Francisco, every weekday at 4:45 a.m. Pacific, 7:45 a.m. Eastern.

[Send me tomorrow’s briefing →](https://www.implicator.ai/subscribe/?utm%5Fsource=newsletter&utm%5Fmedium=cta&utm%5Fcampaign=briefing%5Fsignup&utm%5Fcontent=variant%5Fa)

About five minutes. No hype. No spam.