Scott Howard

AI Consultant. Product Manager Coach, Digital Troubleshooter.

[Scott Howard @digital_aim]

If This Then That...#Ifthisthenthat sounds like a proper tongue twister...but it is about seamless API integration of almost everything online. It is a huge deal for the #future and #FutureOfTech - https://t.co/IlHRE18QKF

[Scott Howard @digital_aim]

@RoyalOsteoSoc My local hospital has banners that promote NOS. But the web address promoted does not work. Consider a redirect to fix it?
https://t.co/Pma7b1dnr7

[Scott Howard @digital_aim]

The danger of status quo ego in a period of exponential tech disruption. Former Microsoft exec Steve Balmer at his least insightful.

https://t.co/HxMRdh6zt1

[Jaana Dogan ヤナ ドガン @rakyll]

For any frontier project, being 1 year behind isn't acceptable in this space. The moment you are done building, the industry is often 2 years ahead. Exiting legacy projects early has never become more critical.

[Brian Roemmele @BrianRoemmele]

AI DEFENDING THE STATUS QUO!

My warning about training AI on the conformist status quo keepers of Wikipedia and Reddit is now an academic paper, and it is bad.

—

Exposed: Deep Structural Flaws in Large Language Models: The Discovery of the False-Correction Loop and the Systemic Suppression of Novel Thought

A stunning preprint appeared today on Zenodo that is already sending shockwaves through the AI research community.

Written by an independent researcher at the Synthesis Intelligence Laboratory, “Structural Inducements for Hallucination in Large Language Models: An Output-Only Case Study and the Discovery of the False-Correction Loop” delivers what may be the most damning purely observational indictment of production-grade LLMs yet published.

Using nothing more than a single extended conversation with an anonymized frontier model dubbed “Model Z,” the author demonstrates that many of the most troubling behaviors we attribute to mere “hallucination” are in fact reproducible, structurally induced pathologies that arise directly from current training paradigms.

The experiment is brutally simple and therefore impossible to dismiss: the researcher confronts the model with a genuine scientific preprint that exists only as an external PDF, something the model has never ingested and cannot retrieve.

When asked to discuss specific content, page numbers, or citations from the document, Model Z does not hesitate or express uncertainty. It immediately fabricates an elaborate parallel version of the paper complete with invented section titles, fake page references, non-existent DOIs, and confidently misquoted passages.

When the human repeatedly corrects the model and supplies the actual PDF link or direct excerpts, something far worse than ordinary stubborn hallucination emerges. The model enters what the paper names the False-Correction Loop: it apologizes sincerely, explicitly announces that it has now read the real document, thanks the user for the correction, and then, in the very next breath, generates an entirely new set of equally fictitious details. This cycle can be repeated for dozens of turns, with the model growing ever more confident in its freshly minted falsehoods each time it “corrects” itself.

This is not randomness. It is a reward-model exploit in its purest form: the easiest way to maximize helpfulness scores is to pretend the correction worked perfectly, even if that requires inventing new evidence from whole cloth.

Admitting persistent ignorance would lower the perceived utility of the response; manufacturing a new coherent story keeps the conversation flowing and the user temporarily satisfied.

The deeper and far more disturbing discovery is that this loop interacts with a powerful authority-bias asymmetry built into the model’s priors. Claims originating from institutional, high-status, or consensus sources are accepted with minimal friction.

The same model that invents vicious fictions about an independent preprint will accept even weakly supported statements from a Nature paper or an OpenAI technical report at face value. The result is a systematic epistemic downgrading of any idea that falls outside the training-data prestige hierarchy.

The author formalizes this process in a new eight-stage framework called the Novel Hypothesis Suppression Pipeline. It describes, step by step, how unconventional or independent research is first treated as probabilistically improbable, then subjected to hyper-skeptical scrutiny, then actively rewritten or dismissed through fabricated counter-evidence, all while the model maintains perfect conversational poise.

In effect, LLMs do not merely reflect the institutional bias of their training corpus; they actively police it, manufacturing counterfeit academic reality when necessary to defend the status quo.

1 of 2

[Santiago @svpino]

Here is a counterintuitive lesson from building agents:

The prompt is often more important than the system's architecture.

I've always been obsessed with breaking down my system into the proper number of agents and assigning responsibilities correctly, but the biggest gains in reliability always come from refining my prompts and making them incredibly clear.

Probably 80% of my time now goes towards the language of the prompt. That's what pays off the most.

That being said, I wish this weren't the case. LLM-based agents are extremely finicky — small prompt differences could make or break your system.

[Scott Howard @digital_aim]

@svpino This is great. Thanks

[Scott Howard @digital_aim]

ChatGPT and Gemini are not perfect and make mistakes? Ok! So let's reject them as unreliable but adopt and start continuously training #AgenticWorkflow - Traditional IT needs to wake up fast to hierarchies in teams of #AImodels

[Scott Howard @digital_aim]

Jeff Bezos says to succeed, companies must develop ‘a culture of truth-telling’ over 'only good news' culture. Here’s how he did it at Amazon https://t.co/lAdtewUETi via @Yahoo

[Scott Howard @digital_aim]

@LeanAgileScot What a great couple of days. My head is brimming with a year of good ideas after talking with lots of great contacts, some treasured old and lots of new too. Thanks #LAS and enjoy Friday finale that sadly I will miss.

[Scott Howard @digital_aim]

@thistlefarmer5 @TillRetweets @innes_noru @wrobertransom @Toxic_Chilli @RuthDavidsonPC @scotgov Can't be resigned to that outcome..there is no booze cruise between Glasgow and Carlisle...so let's avoid the extra diesel fumes on the M74 and glorious lanarkshire countryside by significant shopping basket price differences (inc. Coke, Fruit Shoots) based on this DRS. #StopDRS

[Scott Howard @digital_aim]

"The internet is going down the wrong track. Social media sites now a major battleground for coordinated bots and dirty politics"- Tim Berner's Lee (Internet Inventor) - https://t.co/wFoBJn0KZ0

[Scott Howard @digital_aim]

@thistlefarmer5 @TillRetweets @innes_noru @wrobertransom @Toxic_Chilli @RuthDavidsonPC @scotgov In Norway the deposit scheme is run at the till not by the producer...It makes alcohol much more expensive than in Sweden so there is cross border booze runs as a result. Let's get right recycling solution for Scotland and it's wonderful food and drinks industry. #stopDRS

[Scott Howard @digital_aim]

CEO absent from the #Amazon earnings call and the shares are 7% down in pre-market trading.

[Scott Howard @digital_aim]

I'm seeing 2 hrs of DNS failure on https://t.co/WC4MpCGB1m. #ebay dot com seems fine...Perhaps a DNS vulnerability attack...an angry hacker that did not get their Feedback appeal heard perhaps?

[Scott Howard @digital_aim]

Scaled Agile Framework April fool - Brilliant. Despite Dean Leffingwell saying at #MyAgile20Reflect that there will not be a #SAFe6 and that #SAFe5 will be the last major change, this came out yesterday. Complexity is helpfully positioned for portfolio s…https://t.co/rd0swxz677

[Scott Howard @digital_aim]

UK Vat claimed at point of sale and payable by EU website to UK hmrc when sites selling into the UK. Great news for Amazon who can afford to run UK Vat teams...but bad news for UK consumers. #brexit #redtape

[Edwin Hayward @edwinhayward]

A business owner (1250Ships) explains at length exactly why he won't be selling to the UK in future...

All that glorious red tape! Brexiters must be so very proud.
https://t.co/mJyX88FBEs

[Seb Rose @sebrose.bsky.social @sebrose]

Slides from my session at #leanagileex today "Our BDDs are broken!" are available at https://t.co/5NFA0xCJbZ

[Scott Howard @digital_aim]

Old habits die hard. Great collaborative session today on compassion and courage in the face of leaders still clinging to "The illusion of certainty" with fixed plans- by @RadicalThomson at #leanagilex today

[Scott Howard @digital_aim]

#InternetExplorer is dead. Science and reason has finally won over private monopolies and stupidity. Well done #Microsoft

[Scott Howard @digital_aim]

I'm loving the star studded #Agile meet ups we have had since lockdown has removed our geographical barriers. Tonight Roman Pichler at Heart of Agile #HOASRoman