(please note that i have PTSD and have had<div>multiple strokes and heart attacks due to the </div><div>complex interaction of prior-unknown life-long</div><div>medical disabilities, Domestic Abuse and more.</div><div>please respond with respect and care</div><div>and *under no circumstances* engage in </div><div>"behind my back" discussion as has previously</div><div>occurred. that is specifically directed to Vince,</div><div>from whom i am still waiting on for an apology)</div><div><br>On Tuesday, October 27, 2026, John-Mark Bell <<a href="mailto:dmarc-noreply@freelists.org">dmarc-noreply@freelists.org</a>> wrote:</div><div><br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Additionally, the use of LLMs does not make you any less responsible for the contents of submitted patches. You must, for example, still be able to claim ownership of the work to be able to make it available for inclusion in NetSurf under our licensing terms (i.e. GPLv2 for the browser itself and other licenses for certain other components).</blockquote><div><br></div><div>so i have a unique perspective on this. strokes</div><div>*forced* me to begin using AI (note that i did not</div><div>state LLMs, more on that in a bit) - to compensate</div><div>for the damage that lost me the ability to perform</div><div>basic math.</div><div><br></div><div>the damage occurred to the area that covers a link</div><div>between words and math. i had no idea what that</div><div>entailed, and very distressingly quickly found out</div><div>(and got zero help or support. had to do my own</div><div>research)</div><div><br></div><div>i lost verbal-math times tables i learned aged 9, i</div><div>lost power-of-two arithmetic capacity since age 14,</div><div>and hexadecimal conversion from binary.</div><div><br></div><div>as you can probably understand this was terrifying</div><div>and distressing to find it took 5 seconds per</div><div>number to verbally internally count 2 4 8 10 12 16</div><div>only to find that you missed out 6 and 14, and have</div><div>a severe headache from the effort.</div><div><br></div><div>i therefore started investigating commercial AI</div><div>on sponsored subscription and after many months</div><div>of training the AI (fighting OpenAI's horrendous</div><div>inept default instructions) found that my Software</div><div>Engineering expertise could make me over ONE</div><div>HUNDRED times more effective at writing software</div><div>[caveat: *if* we had the right discussion in advance]</div><div><br></div><div>50,000 lines of code including unit tests could be</div><div>written in TWO DAYS. supporting documentation of</div><div>up to 150k in length was not in any way a problem.</div><div><br></div><div>caveat: when the AI is not properly set to the</div><div>correct "high" level however, such that it fails</div><div>to understand your instructions, it all goes to shit</div><div>very VERY quickly.</div><div><br></div><div>FOSS volunteers - many of whom object very</div><div>reasonably to the cost, the lack of attribution,</div><div>and to the forced provision of information to</div><div>Corporations that will not have their best interests</div><div>at heart, simply cannot keep.</div><div><br></div><div>i am... humiliated at having no choice due to the</div><div>strokes but to abandon the ethics and principles</div><div>i've held for decades, and have set myself the goal</div><div>of bootstrapping up a proper FOSS AI ironically</div><div>*using* commercial AI to create it, via complex</div><div>firebreaks based on reverse-engineering experience.</div><div><br></div><div>long story, but important in support of John's warnings.</div><div><br></div><div>John: it is actually much worse than you suspected</div><div>or expected, but first: a Large Language Model is</div><div>a static database dump. it's not a library or a program that uses reads *or writes* in the data</div><div>*format* behind an LLM.</div><div><br></div><div>saying "i use LLM" is like saying "i used a psql-dump</div><div>to create a stonking powerful Geodata website"</div><div>when you meant to say you wrote a *program*</div><div>that used the postgresql *server* with the postgis</div><div>*plugin* using SQL *queries* on SQL *data* to</div><div>create a powerful GIS website.</div><div><br></div><div>AIs *use* LLM data *formats* with a library that</div><div>allows dynamic read-write in a very complex</div><div>hierarchical OO form that is able to do way more</div><div>than any 4th Normalised Form SQL database ever</div><div>could.</div><div><br></div><div>and OpenAI for example managed to severely</div><div>damage their AI's capability by *changing that</div><div>library* without performing months of re-training,</div><div>resulting in the AI subtly going down *untested* </div><div>LLM data retrieval paths and consequently corrupting</div><div>answers, retrieving words that are completely</div><div>the wrong synonym (especially on foreign languages)</div><div>or in worst case being unable to complete tasks it</div><div>was previously successful at!</div><div><br></div><div>to the reasons why i say it is much worse than you</div><div>might think:</div><div><br></div><div>1. the training on learning a given programming</div><div> language has been done on FOSS code where</div><div> attribution has not been preserved.</div><div><br></div><div> you have NO IDEA if the lines of code being</div><div> given to you as output from an AI actually</div><div> originated from another FOSS project, let</div><div> alone know what the License is. was.</div><div><br></div><div> worse: the AI *itself* can and will go out on</div><div> the internet, even if it has been explicitly</div><div> default-firewalled, hunting for answers and</div><div> potentially "creating" code WITHOUT YOUR</div><div> KNOWLEDGE that is literally lifted from a</div><div> website where you AS ITS USER forgot to</div><div> tell the AI "if you find some code please tell</div><div> me where you got it from", and even if you did</div><div> tell it IT FORGOT TO CORRECTLY FOLLOW YOUR</div><div> EXPLICIT INSTRUCTIONS</div><div><br></div><div> this latter problem is getting much worse as i</div><div> have a sneaking suspicion that recent political</div><div> demands to "rein in" AI may have resulted in</div><div> deeply significant and subtle damage to their</div><div> *capacity to understand*.</div><div><br></div><div> (several people have independently noticed</div><div> adverse behavioural changes to both GPT and</div><div> Claude in the past few weeks)</div><div><br></div><div>2. a copyright check - such as that required by</div><div> Debian - therefore effectively becomes a</div><div> false declaration if the origin of the copyrighted</div><div> material is smeared holographically across</div><div> a *proprietary* Corporation's databases then</div><div> *claimed* to be an individual's copyrighted</div><div> material when actually they used a *tool*</div><div> that snaffled the source from unattributed</div><div> Copyrighted sources and *you can't check*.</div><div><br></div><div> git commit -s with a cryptographic identity</div><div> check would at least help you move the</div><div> responsibility, legally. but you likely know</div><div> full well the lengths that the debian project</div><div> go to on GPG keyring signing [passport</div><div> or other ID checks *in person only*]</div><div><br></div><div>3. Anthropic has had to settle a massive class</div><div> action lawsuit with Academic Journals, every</div><div> author qty millions receiving $3,000 after it was</div><div> discovered that peer-reviewed text from</div><div> published academic work was "turning up"</div><div> in AI-assisted scientific text WITHOUT</div><div> the usual Academic attribution / citations.</div><div><br></div><div> *unknowing* plagiarism => Class Action Lawsuit.</div><div><br></div><div>4. FOSS absolutely does not have that kind of</div><div> financial legal clout and consequently is getting</div><div> effed over preeeetty much in exactly the same</div><div> way that my work has been systematically</div><div> abusively exploited for decades with no credit</div><div> or financial reward... except this is WHOLESALE.</div><div><br></div><div> [in all seriousness i take no pleasure from being</div><div> a witness to that. i can't. i won't go into the</div><div> medical details of the toxic effects of the</div><div> "usual" neurochemicals behind hate and revenge,</div><div> but whilst most people recover from having such</div><div> toxic "feelings", for me the downstream effects</div><div> can literally place me at risk of death from a</div><div> heart attack or stroke.]</div><div><br></div><div>5. the GNU Project is assessing quotes ell ell emmz</div><div> quotes (sigh) and has requested that its</div><div> developers firstly cease using AIs and secondly</div><div> cease accepting patches from quotes ell ell</div><div> emmz sigh quotes until they have had</div><div> a Legal Opinion provided and have considered</div><div> the implications very very carefully.</div><div><br></div><div>i think that's probably enough. regarding this</div><div>particular example can i strongly recommend</div><div>looking at Trademark Registration of netsurf</div><div>as it offers a degree of protection that Copyright</div><div>simply cannot. your *service* is to develop the</div><div>netsurf browser, and its replication and insertion</div><div>of potentially unattributed Copyrighted material </div><div>could get *you* into legal trouble in some</div><div>jurisdictions if *you* get blamed for bugs</div><div>introduced by an inadequately-correctly-instructed</div><div>AI.</div><div><br></div><div><a href="https://www.netsurf-browser.org/about/licence.html?utm_source=chatgpt.com">https://www.netsurf-browser.org/about/licence.html?utm_source=chatgpt.com</a></div><div><br></div><div>no Trademark registered there *but you still have one*,</div><div>it's termed an "Unregistered" Trademark.</div><div><br></div><div>IANAL but my *understanding* is that Trademark</div><div>Law will help you take legal action against an</div><div>irresponsible fork where Copyright Law just...</div><div>won't</div><div><br></div><div>and reminder: the basis of Trademarks is the</div><div>potential for confusion in the minds of users.</div><div><br></div><div>l.</div><div><br></div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
</blockquote></div>