• Welcome to AppraisersForum.com, the premier online  community for the discussion of real estate appraisal. Register a free account to be able to post and unlock additional forums and features.

Anthropic's Claude.AI and Claude Code Status.

RCA

Elite Member
Gold Supporting Member
Joined
Jun 27, 2017
Professional Status
Certified General Appraiser
State
California
Must be the first time today that I have experienced serious downtime for Claude Code. Apparently Anthropic is having a very hard time keeping up with customer demand. Wait until it IPOs? .... Price will skyrocket.

====

I ask Claude: "How come Claude Code is having problems?"


Claude responded: "Let me check whether there's an active incident or known issue with Claude Code right now.

... There's an active outage today. Claude is experiencing a widespread service outage on June 18, 2026, with disruptions starting around midday and complaints spiking sharply, affecting Claude Chat, the website interface, and Claude Code. Users are seeing complete server timeouts, no chat responses, and login errors. So if Claude Code is erroring out or hanging for you right now, you're not alone and it's almost certainly the platform-wide incident rather than anything on your end. CollabnixCollabnix

For the immediate fix: there isn't a client-side one. When it's an infrastructure outage, the official status page is the thing to watch — status.claude.com — rather than reinstalling, re-authing, or fiddling with your config, which won't help and can just add noise.

Worth a bit of caution on the "why," though. A lot of the coverage above is from third-party blogs and aggregators speculating about root cause — sub-agent loops, capacity bottlenecks, and so on — and Anthropic itself had not released an official post-mortem at the time those pieces were written. I'd treat the specific technical explanations as unverified until Anthropic posts its own incident report. The one solid fact is that there's a real, acknowledged multi-platform disruption today. Collabnix

This is also part of a pattern worth noting if Claude Code is load-bearing in your workflow. There have been recurring stability incidents across 2026 — notably March 2, June 2, and June 18 — coinciding with sharp growth in usage. One of the earlier ones had a costly wrinkle for paid users: the June 2 disruption was traced to a bug in Claude Code's sub-agent system, where sub-agents multiplied in an infinite loop and spiked token consumption. If you were running Claude Code around that date, it's worth glancing at your usage to confirm nothing ran away. AimadetoolsAimadetools

The broader takeaway that the engineering-press coverage keeps landing on is one you'll recognize from 40 years of building systems: hard-coding a single provider's endpoint is a single point of failure, and the resilience options are multi-LLM redundancy with automated failover — at the cost of added complexity and the need for a solid eval suite to confirm each model produces acceptable results. For ValEngr specifically, that argues for keeping your API-calling layer model-agnostic enough that you can fall back from Opus to another model without rewiring, which dovetails with the fallback-handling work we were already discussing for Fable. "


I also asked: How many customers does Claude have now?

Claude responded: "Let me search for the most recent figures on Anthropic's customer base.

Here are the most recent figures, though "how many customers" splits into a few different numbers depending on what you're counting.

Business customers: Anthropic serves over 300,000 business customers as of its Series F announcement. That figure has held steady across reporting into mid-2026 — it's the number Anthropic itself cites. For scale, that's up from fewer than 1,000 business customers two years prior. Claude API Docsmexc

High-value enterprise accounts, which is where the growth is most dramatic: the number of customers spending over $100,000 annually has grown roughly 7x in the past year. At the top end, the $1M+ tier moved fast — over 1,000 customers now spend more than $1 million annually, having doubled from 500+ in under two months as of April 2026, up from about a dozen two years ago.

Individual users (the consumer side, harder to pin down since Anthropic doesn't publish a single official number): third-party estimates put it around 18.9 million monthly active users on the web, plus additional millions on mobile. The app specifically was estimated at 12.48 million monthly active users in February 2026, up about 49% month over month.

A couple of caveats worth keeping in mind, since you work with this kind of data:

The 300,000 "business customers" figure is the one Anthropic states directly, so it's the most reliable. The MAU numbers are third-party analytics estimates (Similarweb, AICPB, and similar), not official disclosures, so treat them as directional. And one all-in figure you'll see floating around — a claim of close to 300 million total users including indirect API users — is a loose estimate that mixes direct and indirect usage, so I'd be skeptical of it as a precise count. TechRadar

Also worth noting given the API work you're doing: enterprise and startup API calls drive the majority of revenue through pay-per-token pricing, with business customers accounting for roughly 80% of revenue. So as a solo S-Corp API user, you're part of that 300,000+ business-customer count rather than the consumer MAU figures — the long tail of API developers is folded into that business number."

How long is this downtime going to last? .....

 
Claude is back up, and it looks like it is going to get me to 99.6%+ conformance on a set of rather extensive Prolog and CLP tests.

I did move away from AMZI to Trealla - and that was a good move. Trealla has CLP(Z) and CLP(B) (integer and boolean constraint logic prolog). I am going to add CLP(Q) and CLP(R) for rational and real numbers. [I think I can do that.]... But I only have so much time to do that, and the Prolog engine I have is certainly good enough for now. It boils down to doing it the fast way, the GPL-2 license, or the slow way through an implementation of Holzbauer specs. If I avoid GPL - I can market my prolog --- which may happen. I don't have a need to make the actual code implementation of Prolog open source; in fact, I don't want to be dragged down by supporting some open source implementation of that, whether I charge for it or not.


Here you can see the advantage of Trealla for CLP: https://www.complang.tuwien.ac.at/ulrich/iso-prolog/phrase (aka Neurmarkel's Conformance Suite)

Claude is going to add unit tests for Neumarkel's Conformance Suite. ... Yada yada yada and no one is interested here, I am sure. Anyway ...
 
Just finished testing on my vProlog interpreter/compiler. Understanding it is built on the core Trealla code which is provided as open source with an MIT license.

1. Treally is Linux, and changes were needed to port it to Windows 11. I also runs on MacOS.
2. CLP(Q) and CLP(R) have been added, as an implementation of Holzbauer's cap(q,r) incremental simplex and other algorithms.
3. The MLS DCG grammar for public and agent remarks was added.
4. More additions are coming.
5. It is geared to run from inside of R or Shiny as embeddable code, and an be installed in a CRAN package without problems.

vProlog is bytecode like CPython (the typical Python executable code).

1781860698046.png
 
Find a Real Estate Appraiser - Enter Zip Code

Copyright © 2000-, AppraisersForum.com, All Rights Reserved
AppraisersForum.com is proudly hosted by the folks at
AppraiserSites.com
Back
Top