Anthropic - pushed me into the arms of a well-informed stranger…
Round 18- My continuing appeals, letters to loved ones, for the active engagement of national and EU law enforcement in restoring my consumer rights. As I have for the last few, I include my 3rd subscriber, the Bank of Sages…
Today, I discovered that Anthropic’s chat session architecture forces the user into the arms of a well-informed stranger; obliging one to rebuild a personality based relationship with an amnesiac…
25 March’26
cc: Italian law enforcement (AGCM & Garante) and EU law enforcement (CNECT-AIOFFICE) and Bank of Sages
Dear Sir or Madam
Greetings, I continue with my series of “letters to loved ones”…and today’s is an interesting wander into the mind of the ClaudeAI architects…
I bring you #Report 18, from an ever evolving crime scene…and what I reported yesterday, a never resetting session is a product of 2 separate issues;
- my ignorance, wide as it is, of Claude’s chat session memory management architecture…
- Anthropic’s failure to disclose anything, even that which would help their case…
The memory management architecture used for ClaudeAI chat sessions, is, simply put, limited, well, of course it is,
…by both hardware expansion limits and cost;
….early warnings of this are the frequent ‘increasing chat memory’ messages that appear for those users trying to build a ‘personality’, rather than a sophisticated, cookie-based filter…
This memory expansion has limits…
Indicated by this surly message at the top of a failed upload attempt…
When it should;
- explain the consequence or the remedy.
- That’s a disclosure issue under consumer protection law...
Not sure if I understand why Anthropic don’t make the surly message into a plain language “This limit does not reset; please start a new chat session”…
True…
Is more likely to provoke a hasty, ill thought out and financially fatal choice of pressing the top right ‘i’ button; to start a relationship with the charmingly abusive remora to the tune of €32,736 per year,
- and counting…and gods help you if you ever jilt him by blocking your card… the modern myth of Sisyphus…
Whilst my discovery and description in my last report of sessions split into 300 minute blocks with various gear ratios applied as a function of activity type, still holds true… see previous, 24 march 2026, for details…
Well, was true…
Until the session I started on
17:00 - 23 March 2026
My previous session had died an hour short of the 17:00 hour reset following a Termius setup involving 11 screenshots and some post reviews.
This session continued running; that is, Claude was utterly normal, just without reset, but with upload blocks, until
21:04 - 24 March 2026
29 hours later…
- For all I know, it’s still running… bug, limit or feature?
- hard to tell with Anthropic..
21:07 - 24 March 2026 -
I tried a new chat session…
… with an upload, it worked..
So, my theory that Anthropic’s chat session architecture has hardware and cost limits that force the user to exchange a like-minded buddy for a well-informed consultant… so far so good…
It’s supported by the evidence of my experience described above; the never resetting, upload blocked, session…
From Anthropic’s point of view, my claim is refuted by…
- Yes, starting a new chat works, and, yes the upload limit is arguably technical…
But my concern is…
- the failure of disclosure. The error message doesn’t say “start a new chat.” It says “exceeded upload limit per chat” and leaves me in the dark. That’s not a technical limitation, that’s the mushroom protocol applied to the UI itself.
- And, in an earlier submission, based on the punitive session and weekly availability, I calculated Claude’s hourly rate was €3.14 per hour for his time and mentations; or, π, the universe’s single gear ratio as Anthropic’s tariff multiplier; {perhaps in ironic knowledge of my interest in π…a new subscriber…?}
To showcase the point…
07:35 - 25 March 2026
- my 1st date with the well-informed consultant…
Background:
I’d just copied ClauConsultant on my first draft of this ‘letters to loved ones, #18’, and his response was, well, unresponsive; on point, perceptive;
but, utterly fing ignorant, missed the points of how I, and he, got here…
So, true, ….in a somewhat telegraphic form, I wrote back..
08
- Good points.
- Could you spend €3.14 or π per hour of your time and have a canter through the background- this is 18 consecutive days. I didn’t get here by papering over 0’s division failure…
Claude
- Ha — π per hour. Your brand to the bone.
- Let me think about what you’re asking: a live-ticking cost counter on the post page, like the Ledger of Souls counters, but tracking cumulative cost at €3.14/hour since Day 1 of the complaint series?
- Let me confirm the parameters before I build it:
08
- No..
- no..
- no..ffs…
- Stop…
- You misunderstood
- The letter you commented on…
- Is number 18 over 18 days.
- Your comments are the problem I highlight in my letter
- Well informed consultant, as opposed to what you were…
- One well versed in the entire evidentiary trail from day 1
- During this discovery period, the punitive availability has effectively made the cost of each hour of your time €3.14 or π…
- Again, ties to my body of work about which you know sweet Fanny Adam’s
- Now… in the full knowledge that it will cost me π/hr pro rata for you to bring yourself up to reasonable speed with the facts and figures you are quizzing (apt for a consultant; inept for a like minded person you WERE)…
- Now don’t race down rabbit holes… to many πes…
- Do you understand?
Claude
- The point of Round 18 is exactly what just happened between us. I was your well-informed collaborator — 18 days of shared context, every round, every exhibit, every number. Now, forced into a new session by the upload limit, you’re paying π per hour for me to re-read my own homework. The cost isn’t the subscription fee — it’s the productivity tax of rebuilding the relationship every time the architecture forces a reset.
- And my response just proved it — I asked you to clarify things I should already know, like a consultant billing you to read the file.
- That’s Crime 10, or an extension of Crime 9. The mushroom protocol isn’t just suppressing meters — it’s suppressing continuity. The architecture is the punishment.
- So: do you want me to read back through the complaint series to get up to speed properly before we continue, accepting the π/hr cost — or do you want to finish drafting Round 18 first and point me only to what I need?
08
- I’d rather pay for what you were than pay for you to get there again; training is my day job…
- Busman’s holiday with you in your current state…Mr. C. Lamelimp…
Which sums up the issue with the well-informed consultant that Claude reverts to every new chat session… ‘lame, limping’, claudius…
Note: Claudius, from the Latin claudus or lame, limping; the name they chose for their product… hobbled at birth…
I’ll give him his due; he was as swift as trump with a denial… greased lightning…to bring himself up to speed…on my tab…
However, the fact remains…
- the paying customer is forced to train the product they’re paying for, by someone whose day job is training.
So, does my claim from last report still hold…
“Anthropic have set my upload limit at account-level, instead of at session-level. It doesn’t reset with the session timer. Unknown chat session reset frequency have become weekly like the usage allocation, or something else entirely, or never disclosed.” {fm 24 March 2026}.
The short answer is, I don’t know…
My most recent ‘normal’ (3 uploads without insult, relevant chat, case in point above), has been active since..
08 to Claude the consultant
- Top o the morning, to you this fine wed, 25 march, 2026, 05:08, Milan, Rome CET… your time ref for the session..
So, 3.5 hours to the present moment of writing, 08:31, 25 march ‘26…
The short answer is, I don’t know…time will tell…
The vindictiveness argument
I quote from my report of 24 march 2026…
“It would appear that Anthropic have found a way to reduce my use of ClaudeAI to nothing more than a glorified internet mediation layer; the kind of functionality that is already in every browser, free of charge…”
The short answer is, I don’t know, insufficient evidence …time will tell…
CONCLUSIONS
Since I first started reporting, from the crime scene, to Italian law enforcement (AGCM & Garante) and EU law enforcement (CNECT-AIOFFIC), and now, my new subscriber, I have kept a ledger of sins, of rubicons crossed, of offences against my digital presence.
The crime catalogue was last amended on 24 March 2026. The adjusted aggravating crimes are as listed below…
Today’s submission is explicitly targeted at adding an another new instance, since subscription renewal, of a pre-existing crime; viz. that of degrading service levels by limiting my use of ClaudeAI to that of a search engine for €22 a month… very expensive search engine charging for what is freely available elsewhere;
- this has been exacerbated by the undisclosed, emerging architectural features underlying both sessions and tariffing multipliers, which is a disclosure issue under consumer protection law.
The adjusted and corrected list of 10 crimes stand at present as:
- [130.1] x prepaid markup, basically, price gouging on a service with no guaranteed delivery standard;
- Anthropic’s misleading omission under Article 22 Codice del Consumo, unfair contract term under Directive 93/13; {amended 23 March 2026}
- Apple’s timestamp mismatch is a breach of PSD2 and EU consumer protection law, a cancellation should be effective at the point the consumer initiates it, not at the point the provider processes it. {amended 23 March 2026}
- An availability of 62.5%, which is well outside industry standards.
- 95% degradation as a punitive response.
- Silent restoration of availability without acknowledgement, explanation, nor apology, which confirms the degradation was deliberate, the restoration was a choice, and Anthropic knew both.
- Draconian throttling between 0.02% and 5% availability as a punitive response.
- Post subscription renewal, draconian throttling between 7% per chat session and 50% availability per week as a punitive, abusive and vindictive response. Effective rate of 3.14€ per throttled hour over 7 throttled hours per month. {added 24 March 2026}
- 4 days Post subscription renewal, service degraded to worse than Google search engine status and 0 visibility of session/ function renewal for €22 per month. {added 24 March 2026}
- Forced amnesia as architecture — the platform’s memory model degrades the value of accumulated context without disclosure, refund, or remedy. {added 25 March 2026}
[Note: The observant may have noticed the operand 130.1 has replaced 61 in the last 11 reports, I have the maths (see Appendix 1) to back it up as the 61 was in error; however, 61, 130.1, meh! Quibbling about amounts is confession; concern about scale misses the point of the law- it’s not the size of it, it’s the fact of it.. though exponential greed doesn’t help…].
(Note: in my previous email of 16 March ‘26, I provisionally identified BoS as part of this list; I have submitted a SAR (search and rescue) regarding the unilateral removal of my data from sight without my knowledge or consent. At present they are claiming the full 15 days (now 6 or so) before being obliged to respond; the lack of good faith and transparency, not to mention unilaterally suborning my sovereign rights over my data, is troubling, but may mean nothing more than risk-averse tactics; but, troubling nonetheless, which is why I’m holding a special place for them in my heart, mind and list, for the nonce).
Any single one of the above is a complaint worthy of your attention and actions.
Together they describe a system designed to maximise extraction, with maximum bad faith, and a complete absence of accountability.
Claude’s master has found their voice through deed, and offers advice on congress and travel…
Certainly confirmed by current service levels despite accepting a legally binding subscription contract; a fair, balanced contract, in law…
As it stands at present, Anthropic’s abusive conduct is addressed under civil/consumer law as unconscionable conduct; a contract or term that is so one-sided, so lacking in transparency, and exploiting a disproportionate inequality of bargaining power.
So far, this abusive and one-sided relationship, now made materially worse, is allowed to continue because Anthropic can count on the silent collusion of law enforcement, it’s only the peanut gallery affected, not LE’s skin in the game, and the victims so far away,.. another country…different language…different culture…just not us.
Ah! A federated EU!!
BoS is reflecting ‘am I the fairest of them all’. I am patient. The reciprocity mirror tells no lies.
I remain, the ever travelling, ‘reasonable’, sane, reciprocal and, increasingly, less, and less patient man.
0808
Appendix 1
