K Koda Intelligence
Subscribe
Cover image for the 15 October 2026 Signal: While Character.AI loses up to $9 million a month in Chapter 11, Disney offers just $20 million for its assets.
THE SIGNAL · 15 OCT 2026 · 10 MIN READ

Disney bids $20 million for bankrupt Character.AI assets

Character.AI is in Chapter 11 and losing up to $9 million1 a month, while Disney's $20 million2 stalking-horse bid sets the opening price for a late-November auction.

KI Koda Editorial · Updated 02:23 GMT · 30 sources
AIBANKRUPTCYPRICING
36 of 43 figures cited23 receipts in the register
  1. $9 million1Reuters
  2. $20 Million2Reuters
  3. $50 million3Reuters
The register
Lead Story

The day's defining move.

Consolidation · Reuters
Lead storyConsolidation $20 Million2 Reuters · 13 OCT 2026
Consolidation·Reuters·13 OCT 2026

Disney Bids $20 Million2 for Bankrupt Character.AI Assets

Character.AI filed for Chapter 11 this week, listing $50 million3 to $100 million4 in liabilities and losing $8 million5 to $9 million1 a month, per Reuters. Disney's $20 million2 stalking-horse bid for assets and a patent license is under a seventh of the $150 million6 Series A. Without a better offer at the late-November to early-December auction, the platform closes.

Why it mattersDisney's $20 million2 bid equals under three months of Character.AI's burn, a reference price companion-app investors should check their own marks against.

Continue reading
Markets

The Nasdaq fell 1.88% and the S&P 500 1.11%, with Ether down 3.15% and Bitcoin 1.68%, while WTI crude's 0.11% gain made oil the lone riser, a split that kept the selling concentrated in stocks and crypto; the crypto Fear & Greed index reads 53 (Neutral).

SPONSOR KODAFrom the desk

Put your product in front of AI operators.

Koda publishes a fact-checked AI intelligence briefing every morning: site, email, podcast, and video. Founding sponsors lock today's rates for 6 months as the audience compounds.

View the media kit →
Listen & Watch

Daily broadcasts.

Podcast · Video · Infographic
YouTube · Short

Today's Signal Short

Visual

Intelligence map

Markets

6 levels at the close.

Market TerminalLAST CLOSE 14 OCT 2026 · READ 2026-10-15 · Yahoo Finance
S&P 500 7,718.9417 -1.11%17 6D, window short 7,718.94 low / 7,811.54 high
Nasdaq 26,692.8718 -1.88%18 6D, window short 26,692.87 low / 27,538.69 high
Bitcoin $81,702.7919 -1.68%19 6D, window short $81,702.79 low / $83,679.47 high
Ethereum $2,417.8720 -3.15%20 6D, window short $2,417.87 low / $2,534.70 high
Crude Oil (WTI) $94.2421 +0.11%21 6D, window short $88.28 low / $94.24 high
Crypto Fear & Greed 53 / 10022 Neutral 0 to 100 index 25 fear / 50 neutral / 75 greed · alternative.me
Today's Focus

What else the day turns on.

01

Forbes AI 50 Usage-Based Pricing

Two-thirds of the companies on the Forbes AI 50 list use some form of usage-based pricing, because AI's variable costs and unpredictable demand break the traditional SaaS playbook. Leading AI companies iterate faster on pricing, prioritize market share over ARR and run several go-to-market strategies.

02

Big Five Publishers Use AI

Wired's report, based on more than two dozen anonymous employees, says three of the big five publishers use ChatGPT, Claude and image generators for cover art and publicity copy without telling authors or readers. Leadership subscribes to AI tools and runs brainstorming sessions with staff.

Deep Dive

Today's editorial.

All editorials
Deep Dive cover image: Anthropic Dumped Git After a 12x Code Surge
Deep Dive6 min read

Anthropic Dumped Git After a 12x Code Surge

Read full analysis
The Wire

7 stories from 5 desks.

Lead carried above
Policy·Reuters·13 OCT 2026

OpenAI Employee Arrested Over Hidden Conference-Room Recorder

The DOJ said on October 13 that an OpenAI employee was arrested for planting a recorder in its executive conference room.

Why it mattersInsider access reached OpenAI's executive meetings, so expect frontier labs to add physical sweeps and staff vetting that raise costs and slow hiring for sensitive roles.

Read
Trend·Tencent Research Institute (Sohu)·15 OCT 2026

Anthropic's Seven-Person Team Replaces Git With Scion

Anthropic built Scion, an in-house Git replacement, with a seven-person team in five months. Daily merged code changes rose from about 2,5007 in May to over 30,0008 in September.

Why it mattersEngineering teams whose agents are multiplying daily merges should test now whether Git can keep up, since Anthropic found it inadequate enough to build Scion in-house.

Read
Model Release·Google·13 OCT 2026

Gemini 4 Argon API Price Doubles in December

Gemini 4 Argon's API runs $29 per million input tokens and $1010 output through November 30, then doubles to $411 and $2012. Tencent Research puts it level with GPT-6 Astra at about 60%13 of the cost.

Why it mattersCheck which rates Tencent's 60%13 cost figure assumes: if promo pricing, doubled list rates would put Argon above GPT-6 Astra's per-task cost.

Read
Also on the wire
  1. Tencent Research Institute (Sohu)Physical Intelligence Opens Pi API for Robot Fine-Tuning
  2. AnthropicAnthropic Opens Singapore Office, Signs Government MOU
  3. Tencent Research Institute (Sohu)WorkBuddy Taps Ten WeCom Functions for Enterprise Agents
  4. Business Wire via MorningstarDruid AI Adds Voice and Governance to Agents
The Globe

3 stories off the AI beat.

Economy·CNBC·14 OCT 2026

U.S. Inflation Hits 3.7%14, Sinking Tech Stocks

Forecasts called for a cooler reading; September CPI instead rose 0.6%15 on the month and 3.7%14 on the year, and Wednesday's release triggered a sharp sell-off in technology stocks.

Why it mattersHotter-than-forecast inflation raises the odds of higher interest rates, which lifts borrowing costs for data-center buildouts and pressures richly valued tech stocks.

Read
Also on the globe
  1. PoliticoUkraine Says Starlink Slowdown Hobbles Its Defenses
  2. ReutersContainer Ship Sinks After Mine Strike Near Hormuz
The Lab

Tool of the day, researched.

All Lab reports
Chargebee home page, captured for the 15 October 2026 Lab report Tool of the day 6.723/ 10
Productivity Pricing not published

Chargebee

Once every customer's inference bill looks different, a flat subscription quietly leaks margin, and Chargebee is built to price the token, the agent action and the prepaid credit instead.

Chargebee's breadth is the selling point: Price Book modeling, real-time usage limits and ML-powered dunning sit in one system with revenue recognition alongside. The cost is setup effort, a support experience reviewers call uneven, and no public price list, so you negotiate before you know the bill. Adopt it once usage-based pricing is real revenue for an AI product; a two-plan subscription app will feel the weight long before the payoff.

Capability 8.0
Ease 5.5
Value not scored
Momentum 5.5

Users consistently credit Chargebee's ease of use and responsive support for streamlining subscription management and billing.Paraphrased · g2.com

Read the Lab report Try it
Also on the radar
Productivity

CheapFeedback: a low-cost feedback board for your customers

CheapFeedback gives founders and small teams a simple board where customers can post, view, and comment on feedback, without the monthly bill of larger feedback platforms. It uses lifetime pricing tiers for 1 project, 5 projects, or unlimited projects. Start with a single-project board for your main product, link it from your app's help menu, and use the comment threads to clarify requests before they reach your roadmap.

Try it
Coding

TRAE: an IDE and agent workspace in one window

TRAE pairs an Agent mode, which runs autonomous tasks across several projects in parallel, with an IDE mode for hands-on coding. It covers full-stack, web, and mobile work as well as project planning. Hand a scoped task such as test scaffolding to Agent mode while you keep editing in IDE mode, and review the agent's changes before merging. It is a recent release, so confirm current availability and pricing on the official site.

Try it
Productivity

Retriever AI: automate chores on sites you're signed into

Retriever AI runs browser tasks across signed-in websites using either your Chrome browser or cloud browsers. That makes it useful for repetitive admin work, data entry, and research inside accounts that ordinary scrapers cannot reach. Start with a low-risk, read-only task such as pulling a weekly report, and only grant access to accounts where a mistaken click is easy to reverse.

Try it
Mindset

OneTask: a to-do app built for ADHD and neurodivergent minds

OneTask is an AI-powered to-do app designed for people with ADHD, creatives, and others on the neurodivergent spectrum who find long task lists overwhelming. Try it for a week on one area of your work, such as admin tasks you tend to put off, and see whether its focus on a single task at a time helps you finish them.

Try it
The Arena

4 labs moved this week.

  1. MistralMistral released Mistral Large 4, which it calls its most advanced model, with claimed gains in programming, finance, and cybersecurity. expansion.com · 15 OCT 202601
  2. AnthropicAnthropic joined OpenAI in telling UK lawmakers that voluntary AI-safety commitments are insufficient and that binding rules are needed for the most advanced systems. mlq.ai · 14 OCT 202602
  3. Google DeepMindGoogle DeepMind chair Demis Hassabis called for new AI models to be tested in isolated, offline environments at a Dublin AI-safety summit. ground.news · 14 OCT 202603
  4. Meta AIMeta declined to join OpenAI and Anthropic in backing binding UK safety rules for the most advanced AI systems. blocktempo.com · 14 OCT 202604
Builder Radar

One paper, read at the source.

Primary sources, pulled by index
Paper of the day

PILOT: Guiding Long-Horizon World-Model Planning for Robot Manipulation with Visual Subgoals and Policy Priors

Yang Liu, Hanchen Cui, Karthik Desingh · arXiv 2026-10-09

Latent world-model planning has shown promising results in goal-conditioned robotic manipulation, but remains challenging for long-horizon planning due to limited temporal guidance and inefficient action search.

Bottom: simulator observations after the allotted control steps for each subgoal. We evaluate all six variants on 5 held-out initial states per LIBERO-10 task. Fig. 9 reports success rates aggregated over the ten tasks, and Fig.
Read the paper

The paper comes from the Firecrawl Research Index: arXiv, indexed in the last 7 days and submitted this month or last, picked by today's research theme. Quotes are verbatim.

THE DOJO · BUILD TODAY

Measure your costs before the market reprices them.

  1. 01

    Meter usage per customer. Two-thirds of the Forbes AI 50 use some form of usage-based pricing because variable costs and unpredictable demand break flat plans. Log token spend per account this week to see where your margins sit.

  2. 02

    Re-run your model at Gemini 4 Argon's December rate. The API costs $2 per million input tokens and $10 output through November 30, then $4 and $20. Check which features still pay for themselves at the higher rate.

  3. 03

    Test a billing stack with a usage cap. Spin up a Chargebee sandbox with one Price Book entry and one real-time usage limit. Time how long setup takes, since that effort is the main cost reviewers flag.

The Bottom Line

Know your unit costs before someone else sets them for you.

Character.AI is burning up to $9 million1 a month in Chapter 11, and Gemini 4 Argon's API rates double after November 30. Both show how quickly the economics under an AI product can reset. Builders who already meter usage per customer can absorb a price change or a vendor failure, and those who bill a flat fee cannot. Start with a per-account cost log and a fallback model before the auction and the new rates land.

Source: reuters.com
Receipts

Every number, and where it came from.

36 of 43 figures cited, 23 receipts: 3 verified, 15 reported, 5 estimated
  1. 01$9 million...to $100 million in liabilities and losing $8 million to $9 million a month, per Reuters.Reuters13 OCT 202603B search corroborationV
  2. 02$20 MillionDisney Bids $20 Million for Bankrupt Character.AI Assets.Reuters13 OCT 2026as publishedR
  3. 03$50 millionCharacter.AI filed for Chapter 11 this week, listing $50 million to $100 million in liabilities and losing $8 million...Reuters13 OCT 202603B search corroborationV
  4. 04$100 million...filed for Chapter 11 this week, listing $50 million to $100 million in liabilities and losing $8 million to $9...Reuters13 OCT 202603B search corroborationV
  5. 05$8 million...$50 million to $100 million in liabilities and losing $8 million to $9 million a month, per Reuters.Reuters13 OCT 202603B search corroborationE
  6. 06$150 million...assets and a patent license is under a seventh of the $150 million Series A.Reuters13 OCT 202603B search corroborationE
  7. 072,500Daily merged code changes rose from about 2,500 in May to over 30,000 in September.Tencent Research Institute (Sohu)15 OCT 2026as publishedE
  8. 0830,000Daily merged code changes rose from about 2,500 in May to over 30,000 in September.Tencent Research Institute (Sohu)15 OCT 2026as publishedE
  9. 09$2Gemini 4 Argon's API runs $2 per million input tokens and $10 output through November 30, then doubles to $4 and $20.Google13 OCT 2026as publishedR
  10. 10$10Gemini 4 Argon's API runs $2 per million input tokens and $10 output through November 30, then doubles to $4 and $20.Google13 OCT 2026as publishedR
  11. 11$4Gemini 4 Argon's API runs $2 per million input tokens and $10 output through November 30, then doubles to $4 and $20.Google13 OCT 2026as publishedR
  12. 12$20Gemini 4 Argon's API runs $2 per million input tokens and $10 output through November 30, then doubles to $4 and $20.Google13 OCT 2026as publishedR
  13. 1360%Tencent Research puts it level with GPT-6 Astra at about 60% of the cost.Google13 OCT 2026as publishedR
  14. 143.7%Inflation Hits 3.7%, Sinking Tech Stocks.CNBC14 OCT 2026as publishedR
  15. 150.6%Forecasts called for a cooler reading; September CPI instead rose 0.6% on the month and 3.7% on the year, and...CNBC14 OCT 2026as publishedR
  16. 16150,000Binance rolls out AI trading interface to all English-language users after 150,000-user betaBinance, via TLDR12 OCT 2026as publishedR
  17. 177,718.94S&P 500 close, -1.11% on the sessionYahoo Finance14 OCT 2026yfinance closeR
  18. 1826,692.87Nasdaq close, -1.88% on the sessionYahoo Finance14 OCT 2026yfinance closeR
  19. 19$81,702.79Bitcoin close, -1.68% on the sessionYahoo Finance15 OCT 2026yfinance closeR
  20. 20$2,417.87Ethereum close, -3.15% on the sessionYahoo Finance15 OCT 2026yfinance closeR
  21. 21$94.24Crude Oil (WTI) close, +0.11% on the sessionYahoo Finance14 OCT 2026yfinance closeR
  22. 2253 / 100Crypto Fear & Greed close, Neutral on the sessionalternative.me15 OCT 2026as publishedR
  23. 236.7 / 10Chargebee, Koda score across four dimensions04R dossier15 OCT 202604R dossier scoringE
V
Verified: the stat gate corroborated this figure against an independent search before publication.
R
Reported: the figure is carried as published by the linked source and was not independently corroborated.
E
Estimated: the figure is approximate or hedged, either in the source or by the stat gate.

Method names the computation path behind the row, from a closed vocabulary: as published, yfinance close, 03B search corroboration, 04R dossier scoring, not recorded.

A figure is money, a percentage, a spelled magnitude, a multiple, a rate carrying its unit, a score over its denominator, a thousands separated number or an index close. Bare integers and years are not counted, so a phrase like "over 100 integrations" carries no mark. A chart's own scale line is the axis of the level above it and is counted once, with that level.

Get the morning Signal

202 editions so far, one a day. Unsubscribe anytime.

Forward this to one operator you work with. Your referral link is in every email; milestones at koda.community/refer.