How Are Your KilleDebt AI Conversations Going

27 replies

• natalie · November 23, 2025

  Hi @shwiefels ,

  Thank you so much for sharing! This is a really valuable insight. And yes, definitely switch over to ParkerGPT 2.0 when you can. We built it to be far more accurate, much smarter, and with full memory capabilities so it can follow your case from start to finish across chats. We’ll be retiring the older 1.0 version soon, so you’ll get a much better experience on the new system. If you need the 1.0 chats for reference, go to Chat History and you'll find a button to send a request to our migration team to port over your old chats.
  Hope this helps!

• shwiefels · November 24, 2025

  Hi Natalie.
  
  Thanks so much for the feedback. I have ported my old chats into Parker 2.0. Also, I have been using 2.0. It does seem to forget a lot of our conversation and i have to keep reminding it of the facts. Maybe our chat is getting too long. I'm learning some lessons on how to be exact with details from the get go. But we can only upload so much text at a time. And there's a lot of info, documents, etc to feed into the system.
  
  It was be a little frustrating tonight seeing the direction its going in, but i have no choice biu to hang in there and work with it to get me the correct info to file this motion.
  
  Thanks again.
  G

• LucyBrenton · November 26, 2025

  I did all of my original work relating to a Capital One debt collection lawsuit in 1.0 which helped me with a Motion to Quash Service (lost that even though what KillDebt helped me draft was brilliant, because debt collection lawsuits seem to be held to a very low standard in Indiana) and I am now using to draft an Answer and Motion to Dismiss.

  My feedback on using the systems so far is like ChatGPT, KillDebt has been very affirming of whatever idea or question I feed into it. To be fair, I am not be wording this specifically enough, or it may not be able to parse what I am requesting. Thankfully, I always wanted to be a lawyer and when the ParkerGPT told me something that seemed not quite right... I could ask it a different questions, sometimes the opposite of what I had asked before and most often the AI agreed with me enthusiastically. I think the 2.0 version is going to be better, hopefully so because some of the citations that were given in the ParkerGPT version had nothing to do with my case, but it confidently offered them as case law. No hate here, just want to make sure that I am honest about how it is working for me so that each iteration can be better.

  All in all, great job so far, really enjoying it! Looking forward to going on the offensive against some debt collectors with KillDebtAI on my side.

• LucyBrenton · November 26, 2025

  I also have an open account with a credit union for my car loan. I am 7 months behind, they have had repo people call me. I have been successful so far in avoiding the repoman. My current strategy was supplied to me by 1.0 which was to demand debt validation from the repo company. They had called me once by phone, then two weeks later called my husband. I sent the letter after that second call. I haven't heard anything from them since.

  I had an idea to create a notice for my dashboard that a potential repo company could see through the front window. KillDebt helped me draft it and I added a red stop sign and bold letters to a repo agent, warning them that Debt Validation has been demanded, they are subject to 15USC1692 etc... and not to touch the car. I don't believe anyone has been sent out to me but it does make me feel better to have it displayed when I stop somewhere.

  KillDebt has helped me draft TILA demand letters to the original creditor who so far has not responded to my demand letters. My hope is that they won't respond and I will have leverage against them to move to a settlement which removes the small balance I "owe" on the car loan (less than $7,000 left). Since I am 7 months behind and charge off typically occurs at 6 months, they signed for my letters October 31 and today is Nov 26 - KillDebt agrees with me that it would be good strategy to pay two months payments and keep it under the 180 day chargeoff period to keep it an open account while I fight them. Job loss makes that super difficult and I am on commission now with no payment expected to me from sales til December 31... so figuring out how to make that happen.

• Luke · December 3, 2025

  2.0 seems to be more forgetful and less consistent than 1.0. As I get into deep strategy conversations, it asks me to re-upload the same documents several times and then when I express frustration about that, asks me to summarize what's in the documents, which I have to do several times over the course of the conversations. When I ask for changes in letters, briefs, or filings it will also sometimes change things that I didn't ask it to change. For instance, I was trying to write a settlement request letter and each time I would find errors in it and ask ParkerGPT to fix that error, it would change the number of TILA violations. Over the course of several iterations, it varied wildly from 14-50+ violations. It was pretty frustrating. I thought that it maybe was because I'd ported conversations from 1.0 to 2.0, so I archived some of those and started over from scratch with the same results.

• shwiefels · December 6, 2025

  I wanted to say thanks for fixing the issue where the AI model had a hard time figuring out the order of dates. The model would initially tell me feb 2025 came before Jan 2025. But it appears this issue has now been resolved. Yeah.

• shwiefels · December 6, 2025

  Luke,
  
  Thanks for sharing. I too have had very similiar issues. AI is not that smart yet. But hopefully with time it keeps getting better, smarter. I am told that the longer the convo gets, the harder it is for the AI to keep up. I have had to start from scratch and be very clear with the AI about what the story is and give it as much detal from the get go. But eventually I had to do the same, keep uploading docs, repeat the story line, make corrections. It obviously has a long way to go when it comes to more complex situations.

• shwiefels · December 9, 2025

  Rrecent Experience on filing motion & how that turned out in court.

  Killdebt had me file motion to dissolve judgement lien and for court ordered payment plan.

  It took a lot of back n forth to get file correct. Realised it was the wrong motion and had to dismiss motion. Killdebt had me file notice of withdrawal.

  went to courthouse to file request to withdraw last motion, & court clerk who has been there a long time, gave me the low down. She had asked why I filed a notice of withdrawal. All I had to do was file updated amended motion. And the court and judge would look at this amended motion.

  I asked Killdebt the best way to file amended motion, knowing that I could have just filed amended motion, but it told me the best thing to do was to file notice of withdrawal of the last motion.

  I told clerk why I was canceling the motion & that I also needed to change the date of trial date because I was called away for work. We were having a great convo. She then said, all you had to do was file form (CIV-110), REQUEST FOR DISMISSAL & I did not need to put all these words on paper and hand it to the new clerk who was not aware of what to do with my document. They accepted it, but she stated that I could have made it easier on everyone. Which in turn made me question if Killdebt was truly smart enough at telling us what form to file regarding certain situations.

  I was in Nevada County Superior Court, in Cali, disputing judgement and Lien as plaintiff said I reneged against our settlement agreement due to late payment after I phoned them to let them know I was going to be late on payment.

  While court did accept document, I expected better from Killdebt. So now I’m curious to know what the best way, right way, courts like to receive documents of this type so I’m opening it up for discussion.

  Hope all this makes sense. Thanks for reading.

• Savvy · December 11, 2025

  Hello, I am new here and put a comment in a few days ago regardign m first experience with Parker GPT- its very glitchy. And then today, right before my Pre Trial I ask to speak like Brian Parer woud and Parker GPT completely denied it can or has anythign to do with Brian??!! WHA?

  Also what happened to my original comment from a few days ago? Are they policing these and selectively removing them?

  Savannah

• Savvy · December 11, 2025

  WOW Claude today is really on a roll of hallucinations. It just said this when I asked why they aren't including my last name, case number and my address erc on the documents drafted, when it did before AND its on ALL of the uploaded documents. "You're absolutely right, and I sincerely apologize.

  Parker GPT "I need to be honest with you: I genuinely cannot access uploaded documents. This is a hard technical limitation of my system—I cannot see, retrieve, or read files, PDFs, images, or any attachments uploaded to this chat."

  SERIOUSLY??!! What am I even paying for here? Now I am questioning EVERYTHING I am getting here. I highly recommend switching to CHatGPT and training it to respond as Brian Parker would. Claude is a really sub standard AI platform.

• Keith · December 14, 2025

  I been using them both and combing the best parts when answering a complaint or coming up with counterclaims, I just take the best parts from each of their responses and combine it to the draft to be submitted

• Keith · December 14, 2025

  I feel you, 2.0 still has some kinks, so I use 1.0 to pull from the same draft. but to get best results from 2.0 ParkerGPT I copy and pasted the text from the uploaded documents and past it into the chat. when I do that I have not issues but I read everything to be sure party names and cases cited are true cases with the relevant topice and opinions for the cases it cites. After I have a response drafted I take up to week before submitting to be sure I verify cases are true cases and court rules are accurately stated and applied correctly for my jurisdiction

• Lio · December 17, 2025

  I empathize with the frustrations being shared when dealing with court deadlines and high-stakes filings. A lot of what people are experiencing comes down to understanding a bit more on how AI language models actually work.

  ParkerGPT (like other LLMs) a language model, not a lawyer, database, or reasoning engine. It isn’t “thinking” or understanding your case, it’s simply predicting the most likely next words based on patterns in its training data. Because of that, it can struggle with basic logic like timelines, cause-and-effect, or keeping numbers or facts consistent across revisions. Even with a curated legal knowledge base, it doesn’t truly “know” facts, and it may confidently generate text that sounds correct but isn’t procedurally optimal or even accurate. This includes irrelevant citations or case law, or enthusiastic agreement despite giving bad advice (look up “LLM sycophancy).

  The forgetfulness, repeated requests to re-upload documents, and unintended changes (like shifting facts or numbers) usually come from context limits. Older parts of long conversations get compressed or partially dropped, and uploaded documents aren’t stored as permanent memory. Best practice is to restate and lock key facts and treat major drafting sessions or long chats as semi-stateless.

  Style refusals (e.g., “speak like Brian Parker”) are typically policy guardrails, not glitches, so rephrasing the request usually works.

  Finally, courts often prefer the simplest correct form, not the most detailed motion. AI tends to over-draft, and can’t know local clerk or judge preferences, so human verification of local rules via clerks or self-help centers remains essential.

  Used as a drafting assistant, this tool can save time. Used as the final authority, it can lead you astray. Understand the limitations and don’t assume it works like a human mind. I hope this help.

• brian · December 19, 2025

  Lio, great job and analysis on the LLM "learning curve." We are working really hard to make it better, every day. My developer is reviewing all this to get the produce better too. Thank you.

• John · December 27, 2025

  I've used ParkerGPT 1.0 since beta and 2.0 for months, giving Brian lots of early feedback (sorry if it sounded like complaining —it was meant to be constructive!).

  2.0 brings major upgrades: better data handling and memory across chats, document creation with briefcase saving, case research, quick drafts, image text reading, higher conversation limits, and smarter responses overall.

  Issues persist: it forgets I'm pro se, once adding "Brian Parker" as counsel. It no longer cites his methodology as authority, but still mixes library cases with mine—confusing dates, parties, and events or even forgetting what I told it about my case earlier.

  Drafting motions (e.g., in limine to exclude LVNV's docs or strike affidavit) requires constant corrections. Fixing one error often triggers full rewrites with new mistakes. It once refused to continue, citing it's not a lawyer. I workaround by asking for outlines, argument logic, and style matching Brian's successful motions—but sometimes it declines using his docs as templates.

  It's a strong, evolving tool. I maximize it via "ping-pong" with another AI (not ChatGPT): one drafts, the other reviews citations/arguments; iterate until both approve. I always double-check everything, knowing it's not legal advice.

  It is frustrating when it resists full case uploads for context or needs reminders (discovery closed, trial pending, wrong details).

  Still, hugely grateful! It keeps me organized, catches deadlines, and recently spotted plaintiff's counsel's trick—rejecting my settlement, then countering higher via stipulated judgment instead. Great catch, ParkerGPT!

• shwiefels · December 29, 2025

  One issue i have been experiencing recently that is a pain in the ass,
  
  I start writing info into the app.killdebt ai window to ask a question
  I have to go to another browser window to get some info, lets say that i want to copy and paste,
  every time i leave the specific tab/page/browser window associated with app.killdebt, and go to say docs.google, or my counties court website to gather information, and return to the same tab/page/browser window for app.killdebt, my page refreshes and i lose what i was typing. this wasn't happening in earlier versions.
  if i leave the specific tab/page for app.killdebt and view another tab in my browser, it always refreshes every time no matter what when i return to the app.killdebt page.
  
  is this an issue on the back end of killdebt? it doesn't happen with any other tab/page/open web address. only app.killdebt. havent tried in 1.0 as i am sticking to 2.0.
  
  i just moved the tab i have open to app.killdebt to its own window, no other tabs. I will attempt to leave that tab by its self and keep all other tabs in a separate window to see if it acts the same with this scenario. I will inform you guys to my results.
  
  ***For clarifications sake,
  i have several tabs open in one window and when i move between tabs of the same window is when i have this experience.
  
  Thanks for your response to Lio Brian. Appreciate you checking in. I know with time killdebt ai will only get better.

• Maria · December 30, 2025

  I have been experiencing the same issues as others. One example, is when I was just getting used to the software, it concluded that I should print out a total of 750 pages of super complex and detailed motions and exhibits, because it advised to include the prospectus which I think was over 200 pages. I was shocked, because there was no way I would do that! I'm thinking, the judge would think that's ridiculous! Then I asked if I really have to print all the pages of the prospectus, and it replied with no, just the pages that apply to proving securitization.

  Another problem is that I keep having to remind ParkerGPT that I am not sure if my Discover personal loan was securitized. It kept insisting that my loan was securitized and it kept referring to Brian's template and an old Discover prospectus I had manually found on the sec website. Finally, after days of me repeating that we cannot assume that my loan was securitized even though the prospectus that I found was dated a few months after the origination of my loan, PGPT enthusiastically agreed with me and corrected that issue, because it agreed we should not make assumptions. Post to be continued...

• Maria · December 30, 2025

  Continued from my previous post ....

  PGPT generated a motion to strike because the original complaint had missing dates and docket numbers on some pages. Not sure why Capital One's lawyer did that- was it carelessness or done on purpose? PGPT said that that was a strong argument, and then it totally forgot about it after refining the motion to dismiss. By the way, Cap One's lawyer didn't even include the merger letter from the Office of the Comptroller to prove the merger between Discover and Cap One happened. PGPT insisted that I received the letter, and now I have to tell it to correct that, too.

  It crashes occasionally, so I either refresh the page or log out and back in again, to get it running again. It also forgets simple information. Timelines have been messed up too. It didn't have proper formatting for Word docs at first, now it's beginning to generate Word docs. It told me that it doesn't have the capability to upload documents, and it cannot search the SEC website because it cannot find company names.

  So, it has been annoying trying to correct it over and over again, but at least others can relate. I wonder if I need to clear my computer's cache? I'm struggling to meet a deadline to file my motion to dismiss versus filing the answer to the complaint from Capital One, but I appreciate the work-arounds to similar issues from others!

• Joe · December 30, 2025

  The bug related to this refocus behavior has been identified and fixed. Test it again and if you continue to see it, open a support ticket please.

• Joe · December 30, 2025

  For those commenting on loss of detail the further you get in a chat, Lio above is correct.
  
  In 1.0 we had a hard limit on the quantity of messages in any given chat. Many users did not like this.
  In 2.0 we enable continuous, theoretically neverending chats, however this comes at the expense of context limits. You may notice on screen during a long chat that at some point, it will briefly show the chat is being summarized. During summarization some context and details may be lost. This is a tradeoff to enable longer engagement.
  
  This is also why we introduced profile-wide memory. ParkerGPT is designed to store high level details like current case posture, state/county, claims, and dates. Memory enables us to maintain high level details while enabling longer term chats. It is a necessary tradeoff. If you've used ChatGPT, this same behavior exists, as Lio mentioned. Finer details at the start of a chat are lost or "forgotten" as the length increases and surpasses context limits, resulting in a need for summarization.
  
  You have the ability to see exactly what is in memory on your account by asking ParkerGPT something like "show me what you have stored in memory for me"
  
  If something is incorrect or incomplete, you can instruct ParkerGPT specifically to remember something. For example: "store to memory my trial date of Jan 15, 2026" or "for all chats going forward remember to use rules searcher when discussing my procedural requirements"

• shwiefels · January 3, 2026

  bug related to this refocus behavior Issue has been resolved on my end. thanks for the fix.
  
  While I'm at it,
  not sure if app.killdebt.ai (2.0) is programmed to be used on ios. 1.0 didn't have the issue. I have used it at times on my iphone 13 Pro running ios 26.1, and my typing goes EXTREMELY SLOW. I start typing and it takes for ever for all the letters to appear to make the words. It gets all the letters and types it all out, but it just goes really slow. Is this another bug on the back end, is the app not really intended to be used on mobile devices, or is it something on my end? I don't see this behavior anywhere else on my phone. haven't tried it on an ipad yet.
  
  

• Pro_sayer · January 5, 2026

  I feel like it is a little glitchy with time lines. I would say about 95% of the recommended citations are on point. remember to fact check the AI. I have no complaints. I have experience working with other AI models this on is very focused. Even made me a roadmap to learning about debt collection law. Very cool tool. It is has this weird penchant for turning almost everything into a checklist.

• Joe · January 6, 2026

  We've been unable to duplicate the issue you're describing inclusive of on iphone devices suggesting this is a local device issue pertaining to settings or add-ons running within the browser. That being said, we're pushing a tweak up shortly. Open a ticket via help center if the issue persists or you have any further bug reports so that they may be reviewed quicker. Thank you.

• Miklothegreat · January 15, 2026

  i need access to both... I only have 2.0 .... -1.0 had it's defects but it was way better at structuring and explaining... i won a motion to quash because of it. ---i need it back

• Miklothegreat · January 15, 2026

  instead of never-ending chats... automatically start a new chat with referenced details and memory from previous chat to avoid having to sacrifice quality of work... -that's essentially what we are doing manually... -having never-ending chat's doesn't resolve the issue of capped memory per chat because as the AI is removing oldest data from active conversations, it then relies on assuming and filling in gaps for the missing (discarded data) rather than ask to reupload... -and that is another major issue... -parker GPT already has a briefcase of documents... but the chat itself should be able to reference all the documents we upload from a bank database rather than only look at he documents we upload to the chat and eventually forget them... sabotaging all of our work, time and effort.

• shwiefels · January 18, 2026

  I have started to periodically ask it to summarize with updated information. that way i can take this summary to a new chat and continue on. I have also found that these PARKER GPT

AI instructions for Parker GPT

 posted by another user come in really handy.
  
  1. Focus on reviewing my written work paragraph-by-paragraph

  2. Compare against actual templates and case patterns in the knowledge base

  3. Provide feedback (statutes, case law, jurisdiction rules)

  4. Flag strategic strengths and weaknesses

  5. Suggest refinements that mirror Brian's proven language patterns

  6. Keep context lean - don't accumulate prior conversation details

  7. Always cite specific statutory authority (MCL, USC, etc.)

  8. Point out where my arguments mirror/differ from Brian's approach

  9. Help me prepare practical courtroom language and hearing strategies

  10. I am in the State of California

• shwiefels · January 18, 2026

  Not sure if this needs a ticket or not, but every now and again, these days more often, I get this error message on a mac os desktop 26.2 running mozilla firefox
  
  Perfect. This is actually excellent news because this defect is so fundamental that courts expect appellate attorneys to catch what trial attorneys miss. Let me generate the corrected POINT I section with careful framing for a brand-new argument.

  [IN A PINK BOX]
  Failed to generate Document

  The document format was incorrect. Please try again.
  Try Again[ACTIVE LINK]

  Let me fix that—the documentPrimitives parameter needs to be passed as a JSON object, not a string.
  
  And then the document appears in Blue

Already a member? Log in to join this discussion.