K Koda Intelligence
Subscribe

The Signal · Thursday 19 November 2026

Anthropic prices its IPO and opens its books to the public

Anthropic set its IPO price at $83 per share, implying a valuation above $1.4 trillion. Cheaper coding models from Microsoft and a wide supply-chain breach give builders plenty to act on.

Koda editorial team · 19 November 2026 · 6 min read

Listen · 19 min
Cover image for the 19 November 2026 Signal: Anthropic prices its IPO and opens its books to the public

The lead

Consolidation · Reuters · 18 NOV 2026

Anthropic, PBC priced its initial public offering of Class A common stock, the company announced.

Why it matters. Anthropic's customers will gain regular public disclosure of its finances, making its pricing power and long-term stability easier to judge before signing multiyear contracts.

Go further: Anthropic IPO: International Retail Brokers and Selling Agents | Claude Help Center (support.claude.com) and today's paper, How Web Research Agents Interpret User Priorities (arXiv, 17 Nov).

Also today

Benchmark · Tech Times · 18 NOV 2026

Independent Tests Find Smaller Astra Cost Savings

Artificial Analysis and Vals AI measured GPT-6.2 Astra's cost per task. They found cuts of roughly 36% and 50% versus GPT-6 Astra1, short of OpenAI's claim of a drop below one-third.

Why it matters. Budget Astra migrations on a 36% saving1, the lower independent figure, until your own per-task logs confirm OpenAI's deeper cut.

Benchmark · News10 · 18 NOV 2026

Visual Cipher Bench Tests Long-Horizon Agent Reasoning

Einsia AI and Navers Lab launched Visual Cipher Bench, scoring agents on rule discovery, logical planning and spatial operations, the same week FrontierHarness Eval v1.1 appeared.

Why it matters. Agent vendors selling visual automation should publish Visual Cipher Bench scores beside FrontierHarness Eval v1.1 results, because short visual QA numbers miss long-horizon planning.

Microsoft AI, via TLDR

Microsoft Ships Seven Low-Cost Models

Microsoft AI released seven low-cost models. Its MAI Code 2, running in GitHub Copilot, reaches 98% of GPT-6.1 Sol's coding quality at about 40% of the price2, making cost the headline pitch for developers.

Google, via Superhuman

Google widened access to Gemini Tasks

Google widened access to Gemini Tasks, a background mode that works through to-do lists across a user's computer and Google apps. It can handle phone calls and payments and is powered by Gemini 4 Hydrogen across most plans.

theregister.com, via TLDR

Optimizely Breach Reaches 178 Organizations

A breach at Optimizely spread malicious JavaScript across the websites of at least 178 organizations. The count is a floor, so the full number of affected sites remains unknown.

Diplomacy · Reuters · 18 NOV 2026

Trump, Xi and Putin Share Shenzhen Stage

The three leaders had never met as a trio; on Wednesday they appeared together in Shenzhen before trilateral APEC talks. The Kremlin says no full Trump-Putin bilateral negotiations are planned.

Why it matters. With the Kremlin ruling out full Trump-Putin bilateral negotiations, the trilateral APEC session becomes the main place to watch for trade signals that could shift tariff and supply-chain plans.

In brief

Markets

S&P 500 -0.23%, the Nasdaq +0.42% and WTI crude -4.84% on the 18 Nov session. Bitcoin -2.75% and Ether -2.29% on the 19 Nov session. The crypto Fear & Greed index reads 10, Extreme Fear.

S&P 500

7,323.63

-0.23%

Nasdaq

24,429.88

+0.42%

Bitcoin

$63,107.15

-2.75%

Ethereum

$1,781.25

-2.29%

Crude Oil (WTI)

$81.85

-4.84%

Crypto Fear & Greed

10 / 100

Extreme Fear

Levels from Yahoo Finance3, 18 Nov and 19 Nov; alternative.me4, 19 Nov.

Go deeper

The Koda Signal, 19 November 2026

Anthropic prices its IPO and opens its books to the public

0:0018:36

Every episode

Tool of the day

Leonardo AI

Design · Koda score5 7.2 / 10

Leonardo's case for game art rests on repeatability: G2 reviewers credit reusing a proven model and setup for holding one look across many images, and personal model training turns that habit into a feature. Indie developers blocking out props and environments early should adopt it for that phase, starting free and moving to Premium when volume climbs. The bet weakens at the final-asset stage, where small details still need a human hand.

Capability 7.5
Ease 8.0
Value 7.5
Momentum 5.5

Also worth a look

The bottom line

Public filings will make model pricing easier to compare.

Anthropic's move to public markets arrives the same week Microsoft ships a coding model that undercuts the leaders on price, and the two together reward teams that can swap models quickly. Keep your prompts, evals and routing portable so a better price can be tested in an afternoon. Run your own cost-per-task numbers on a sample of real work. A vendor's headline savings are only a starting point for that test.

  1. Benchmark MAI Code 2 on your own work. Pull ten real tasks from your repo and run them through MAI Code 2 in GitHub Copilot and your current model. Record cost and quality for each before trusting the 98% figure2.
  2. Inventory every third-party script. A breach at Optimizely put malicious JavaScript on at least 178 organizations' sites, and that count is a floor. List the external scripts your pages load, then add a content security policy and integrity checks where you can.
  3. Try the Azure DevOps path end to end. Use the GitHub Copilot app for Azure DevOps on a small, low-risk repo, from clone to pull request. Note where you had to step in so you know what to review by hand.

Sources and checks

Every story links to its outlet. Below, by source, are the 11 figures we checked: 2 verified, 7 reported and 2 estimated.

  1. 1Tech Times. 2 figures on this page, 18 Nov 2026.Verified
  2. 2Microsoft AI, via TLDR. 2 figures on this page, 18 Nov 2026.ReportedEstimated
  3. 3Yahoo Finance. 5 figures on this page, 18 Nov 2026 and 19 Nov 2026.Reported
  4. 4alternative.me. Crypto Fear & Greed close, Extreme Fear on the session, 19 Nov 2026.Reported
  5. 5Koda Lab report (our score). Leonardo AI, Koda score across four dimensions, 19 Nov 2026.Estimated

Verified confirmed by an independent searchReported as published by the named sourceEstimated approximate, in the source or in our check

Every figure, one by one (11)
  1. 36%They found cuts of roughly 36% and 50% versus GPT-6 Astra, short of OpenAI's claim of a drop below one-third.Tech TimesVerified
  2. 50%They found cuts of roughly 36% and 50% versus GPT-6 Astra, short of OpenAI's claim of a drop below one-third.Tech TimesVerified
  3. 98%...seven low-cost models; MAI Code 2 in GitHub Copilot hits 98% of GPT-6.1 Sol's coding quality at about 40% of the priceMicrosoft AI, via TLDRReported
  4. 40%...Copilot hits 98% of GPT-6.1 Sol's coding quality at about 40% of the priceMicrosoft AI, via TLDREstimated
  5. 7,323.63S&P 500 close, -0.23% on the sessionYahoo FinanceReported
  6. 24,429.88Nasdaq close, +0.42% on the sessionYahoo FinanceReported
  7. $63,107.15Bitcoin close, -2.75% on the sessionYahoo FinanceReported
  8. $1,781.25Ethereum close, -2.29% on the sessionYahoo FinanceReported
  9. $81.85Crude Oil (WTI) close, -4.84% on the sessionYahoo FinanceReported
  10. 10 / 100Crypto Fear & Greed close, Extreme Fear on the sessionalternative.meReported
  11. 7.2 / 10Leonardo AI, Koda score across four dimensionsKoda Lab report (our score)Estimated

Signal 237 · 19 November 2026 · 14 sources · How every number is checked · Corrections

Get the morning Signal

236 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.