On Wednesday, OpenAI announced ChatGPT Health, a dedicated section of the AI chatbot designed for "health and wellness conversations" intended to connect a user's health and medical records to the chatbot in a secure way.

But mixing generative AI technology like ChatGPT with health advice or analysis of any kind has been a controversial idea since the launch of the service in late 2022. Just days ago, SFGate published an investigation detailing how a 19-year-old California man died of a drug overdose in May 2025 after 18 months of seeking recreational drug advice from ChatGPT. It's a telling example of what can go wrong when chatbot guardrails fail during long conversations and people follow erroneous AI guidance.

Despite the known accuracy issues with AI chatbots, OpenAI's new Health feature will allow users to connect medical records and wellness apps like Apple Health and MyFitnessPal so that ChatGPT can provide personalized health responses like summarizing care instructions, preparing for doctor appointments, and understanding test results.

OpenAI says more than 230 million people ask health questions on ChatGPT each week, making it one of the chatbot's most common use cases. The company worked with more than 260 physicians over two years to develop ChatGPT Health and says conversations in the new section will not be used to train its AI models.

"ChatGPT Health is another step toward turning ChatGPT into a personal super-assistant that can support you with information and tools to achieve your goals across any part of your life," wrote Fidji Simo, OpenAI's CEO of applications, in a blog post.

But despite OpenAI's talk of supporting health goals, the company's terms of service directly state that ChatGPT and other OpenAI services "are not intended for use in the diagnosis or treatment of any health condition."

It appears that policy is not changing with ChatGPT Health. OpenAI writes in its announcement, "Health is designed to support, not replace, medical care. It is not intended for diagnosis or treatment. Instead, it helps you navigate everyday questions and understand patterns over time—not just moments of illness—so you can feel more informed and prepared for important medical conversations."

A cautionary tale

The SFGate report on Sam Nelson's death illustrates why maintaining that disclaimer legally matters. According to chat logs reviewed by the publication, Nelson first asked ChatGPT about recreational drug dosing in November 2023. The AI assistant initially refused and directed him to health care professionals. But over 18 months of conversations, ChatGPT's responses reportedly shifted. Eventually, the chatbot told him things like "Hell yes—let's go full trippy mode" and recommended he double his cough syrup intake. His mother found him dead from an overdose the day after he began addiction treatment.

While Nelson's case did not involve the analysis of doctor-sanctioned health care instructions like the type ChatGPT Health will link to, his case is not unique, as many people have been misled by chatbots that provide inaccurate information or encourage dangerous behavior, as we have covered in the past.

That's because AI language models can easily confabulate, generating plausible but false information in a way that makes it difficult for some users to distinguish fact from fiction. The AI models that services like ChatGPT use statistical relationships in training data (like the text from books, YouTube transcripts, and websites) to produce plausible responses rather than necessarily accurate ones. Moreover, ChatGPT's outputs can vary widely depending on who is using the chatbot and what has previously taken place in the user's chat history (including notes about previous chats).

Then there's the issue of unreliable training data, which companies like OpenAI use to create the models. Fundamentally, all major AI language models rely on information pulled from sources of information collected online. Rob Eleveld of the AI regulatory watchdog Transparency Coalition told SFGate: "There is zero chance, zero chance, that the foundational models can ever be safe on this stuff. Because what they sucked in there is everything on the Internet. And everything on the Internet is all sorts of completely false crap."

So when summarizing a medical report or analyzing a test result, ChatGPT could make a mistake that the user, not being trained in medicine, would not be able to spot.

Even with these hazards, it's likely that the quality of health-related chats with the AI bot can vary dramatically between users because ChatGPT's output partially mirrors the style and tone of what users feed into the system. For example, anecdotally, some users claim to find ChatGPT useful for medical issues, though some successes for a few users who know how to navigate the bot's hazards do not necessarily mean that relying on a chatbot for medical analysis is wise for the general public. That's doubly true in the absence of government regulation and safety testing.

In a statement to SFGate, OpenAI spokesperson Kayla Wood called Nelson's death "a heartbreaking situation" and said the company's models are designed to respond to sensitive questions "with care."

ChatGPT Health is rolling out to a waitlist of US users, with broader access planned in the coming weeks.

Read full article

Comments

Make peace (and progress) with incomplete knowledge

You are always working with incomplete knowledge. If you’re working on something for the first time, there will come a point in future — perhaps tomorrow, perhaps next year — when you find a better way. You may decide against the workaround you’re putting in place now. You may find a simpler way to code something. You may know exactly where to find the problem you’re hunting down. In short, in the future, you know more.

That horizon makes it sound like we can never do anything — or at least anything good — in the present. It sounds like an invitation to overthink, procrastinate or give up in despair. But this breach is where humans excel. We are always working with incomplete knowledge. We start the day without knowing what the day will truly bring — knowledge we won’t have until the hindsight of day’s end — and we get through. Although far from perfect, we’re surprisingly good at it. The issue is not so much that we don’t know everything we need to know, but that we don’t realise it — or that we fool ourselves into believing we do.

It is impossible for one to learn that which one already thinks one knows.

— Epictetus

You can always make progress by trying something out. By trying things out you learn something, you acquire knowledge. By trying things out, you might avoid dead ends, needless detours and treacherous routes. By trying things out you might have something that works for now and, in future, can be improved, built on or replaced based on understanding rather than guesswork or panic.

But don’t forget to come back to it in future. Legacy codebases are mired in the regret of unapplied knowledge. Deferment to the future is a form of commitment rather than one of abandonment, as is so commonly practised.

We should be inspired by compassion rather than blame when we find ourselves frustrated by legacy code. The authors of the code — yourself included — most likely lacked many insights from their future we now have in our present. The inevitability of imperfection doesn’t make whatever problem we’re dealing with go away, but it should temper our frustration and make us less ready to misattribute intention.

There are more things in Heaven and Earth, Horatio, than are dreamt of in your philosophy.

— William Shakespeare

Although we might identify technical debt as being the defining problem of legacy, a simple but significant omission makes this characterisation misleading: technical debt is not intrinsically a problem; unmanaged technical debt, however, is. If you’re managing it, well done! A debt is an obligation to repay and, if you’re fulfilling that obligation, that code is probably not considered with the negativity ‘legacy’ often entails.

So, if your debt is unmanaged, ask yourself why is it unmanaged? Characterising a problem simply as technical debt misses not only an important nuance — managed versus unmanaged — but also the distinction between cause and effect. Technical debt is often blamed, but is an effect rather than a cause. The question to ask is “What caused this?” The answer is typically technical neglect.

The causality of neglect is hard to see because it is an absence of a thing — care, attention, remediation, work — rather than a thing that is easily observed. Inaction lacks the visibility of action. Searching for improvement becomes more a question of “What are we not doing?” than “What are we doing wrong?”

And questions are important—often more so than answers. They help us interrogate what we know, what we believe and what we want — and to distinguish between knowledge, belief and desire. The knowledge you have at the moment is incomplete and imperfect — and, for any given moment, always will be — but your ability to acquire knowledge and put it into practice is ever-present.

None of our beliefs are quite true; all have at least a penumbra of vagueness and error.

— Bertrand Russell

Software development (and life) is far from being a game of complete or perfect information, but that doesn’t mean you can’t play — or that you can’t play well.

On the origins of ‘clean code’

A recent online discussion critiquing the oversimplification, polarisation and moralisation associated with the label clean code repeated a common misconception around the term, namely that it was coined by Robert Martin. He popularised it and associated it with a specific approach in his Clean Code book, but he is not the originator of the term. Although absent from his writings before that book, it was certainly in use by many others long before Clean Code was published in 2008.

Prior to becoming considered a brand or a dogma, cleanliness and cleaning of code was a suggestive and general metaphor rather than a specific checklist of practices and judgements. Talking about ‘clean code’ was more along the lines of ‘clean clothes’ or ‘spring clean’ than ‘clean eating’ or ‘clean desk policy’, with their justifiably negative associations. The term clean code was no more concretely defined or rule-based than, say, tidy code.

For example, we find it in the opening sentences of Kent Beck’s 2003 Test-Driven Development book:

Clean code that works, in Ron Jeffries’ pithy phrase, is the goal of Test-Driven Development (TDD). Clean code that works is a worthwhile goal for a whole bunch of reasons.

We also find it in Martin Fowler’s 1999 Refactoring book:

I’ve not succeeded in pinning down the real birth of the term refactoring. Good programmers certainly have spent at least some time cleaning up their code. They do this because they have learned that clean code is easier to change than complex and messy code, and good programmers know that they rarely write clean code the first time around.

With the metaphor being used extensively throughout, for example:

The compiler doesn’t care whether the code is ugly or clean. But when we change the system, there is a human involved, and humans do care.

Use of this metaphor, however, is not confined to XP practitioners or the software craft lexicon. For example, in 2011, The New York Times’ obituary for Dennis Ritchie:

Colleagues who worked with Mr. Ritchie were struck by his code — meticulous, clean and concise.

But predating any of the uses I’ve already mentioned is what I believe was my first encounter with cleanliness in code:

The best documentation for a computer program is a clean structure.

And clean code:

We will say it once more — clean code is easier to maintain.

These quotes are from Brian W Kernighan and P J Plauger’s 1978 book, The Elements of Programming Style (2nd edition). Although already dated when I first read it — examples are in FORTRAN 66 and PL/I — this book had a profound influence on how I thought about code — formatting, meaningful names, comments, boundary and error conditions, logic and control flow, regularity, refactoring and more. I borrowed it during my first job after university and bought my own copy a few years later.

You can also find these quotes in the first edition, which was published in 1974. The idea of clean code has been around in one form or another for at least half a century.

About 20 new Colorado laws take effect Jan. 1,  doing everything from protecting wild bison to streamlining marijuana regulations. 

We’ve rounded up a few of the most notable new laws here. 

Right-to-repair electronics 

This story was produced as part of the Colorado Capitol News Alliance. It first appeared at cpr.org.

Coloradans will have more options for repairing broken cellphones, computers and other electronic devices starting Jan 1.

A new law requires manufacturers, such as Samsung and Apple, to provide “documentation, software, data and other tools” to device owners and independent repair shops to help people fix their electronic equipment.

The goal is to help consumers save money and get repairs done faster.

The law includes some exemptions, including for video game consoles due to piracy and security concerns. 

Democratic state Rep. Brianna Titone of Arvada was the bill’s main sponsor and says it also applies to sales between businesses and with the government.

“This will save companies a ton of money because companies will be able to hire their own in-house people if they want, or a third-party service provider to do the work of fixing their equipment,” she said. “And they don’t have to go have the expensive contract that the manufacturer requires them to really have.” 

Colorado’s so-called right-to-repair law will be one of the most expansive in the country.  The state already has a right-to-repair wheelchair law, and a right-to-repair agricultural equipment.

Additional paid family leave for parents of kids needing neonatal care 

Colorado’s paid family leave program, FAMLI, allows workers to receive a significant portion of their pay if they need to take up to 12 weeks off work in a given year for a serious family health or personal issue. It’s a popular program that voters approved by a wide margin at the ballot.

Lawmakers have now expanded it for families with a baby in neonatal intensive care. Those families, under a law taking effect in 2026, can apply for an additional 12 weeks of leave. 

Democratic state Sen. Jeff Bridges was one of the main sponsors, and it was inspired by his own personal experience. His son was in intensive care, which he said was “terrifying and consuming.”   

“We need to make it easier for parents with kids in the NICU,” he said.

The measure largely passed along party lines, with opponents worried about increased costs to businesses and workers who pay into the FAMLI program.

Gun show requirements 

Operators of gun shows in Colorado will be required starting in 2026 to submit a security plan to local law enforcement and to hold liability insurance under a new law.

The plans will have to include a list of vendors, a floor plan, information on which areas are under video surveillance,  and the estimated number of attendees.

Gun show operators must also enforce age limits for attendees. Gun shows must also ensure that all purchases follow the state’s background check requirements and the state’s three-day waiting period law.

Screening for renters and fee transparency for consumers

Landlords will not be allowed to ask prospective tenants using a housing subsidy to submit a credit history or credit score as part of the screening process under a new law that takes effect in 2026.

A separate housing bill aims to make prices for products and services more transparent for consumers. The goal is to standardize prices up front so people aren’t hit by hidden fees later on. The law also restricts the kind of fees landlords can charge tenants. 

Protecting wild bison 

Colorado will now classify wild bison as big game wildlife, instead of only as livestock, which will  provide added protections for the animals.

The law, which takes effect in 2026, make it mostly illegal to hunt or poach wild bison, also known as buffalo.

The bill was brought to the state legislature at the request of some tribal communities. The law does not apply to privately owned bison that are in captivity, or bison owned by an American Indian tribe. 

According to Colorado Parks and Wildlife ,the state does not have any confirmed wild bison herds, although sometimes bison from Utah’s Book Cliffs herd cross into the state.

This story was produced by the Capitol News Alliance, a collaboration between KUNC News, Colorado Public Radio, Rocky Mountain PBS and The Colorado Sun, and shared with Rocky Mountain Community Radio and other news organizations across the state. Funding for the Alliance is provided in part by the Corporation for Public Broadcasting.

State oil and gas regulations meant to cut leaks of the highly damaging greenhouse gas methane slashed the unwanted emissions 70% after Colorado launched a series of first-in-the-nation compromises with the industry in the 2010s, says a new study by an environmental group. 

Methane emissions from Colorado’s oil and gas production fell sharply from 2010 to 2017, according to the Environmental Defense Fund study. Researchers used data from the Japanese Greenhouse Gases Observing Satellite, and the findings were backed by results from aircraft-based analysis in flights over the high oil production Denver-Julesburg Basin. 

EDF advocates credit hard-fought collaborations among regulators, oil and gas trade groups and environmentalists beginning with 2014 legislation under then-Gov. John Hickenlooper. Another important methane law passed in 2017, and six more legislative compromises moved forward during Gov. Jared Polis’ administration, EDF officials said. 

“The timing and magnitude of the decline closely align with Colorado’s methane regulations, suggesting they played a central role in driving reductions,” EDF said. The nonprofit advocacy group noted that Colorado’s oil and gas production actually increased during the time that methane emissions dropped, further evidence of the laws’ efficacy. 

Rigorous, carefully negotiated rules like Colorado’s leadership on methane will be more important than ever for continuing gains on greenhouse gases and local pollution as the federal government rolls back mandates under President Donald Trump’s and the GOP’s direction, the EDF said. 

“We are not getting the regulatory protection around methane, greenhouse gasses or climate from the EPA and the federal government at this time, and I think it’s more than ever incumbent on state regulators to take up that that responsibility to protect their residents,” said Nini Gu, regulatory and legislative manager of the EDF’s West region. “And Colorado has always been really great when it comes to oil and gas emissions regulations.” 

Colorado officials and environmental groups say the state’s extensive rules and regulations on cutting greenhouse gas pollution, cleaning up ozone, and promoting clean power will stay in place despite federal rollbacks. Where the federal reversals have threatened Colorado policies, the Polis government and Attorney General Phil Weiser have launched a flurry of lawsuits to keep the status quo.

The new laws in the 2010s emphasized detection of leaks through regular infrared monitoring and other methods, and new leak control equipment. 

Colorado’s industry trade groups welcomed the results of the study, for the same reasons the EDF underlined. 

“Colorado is the nation’s fourth-largest producer of oil and natural gas. This new data confirms that emissions can be reduced while continuing to produce the abundant, reliable and affordable energy that Coloradans rely on,” said Lynn Granger, president of the Colorado Oil and Gas Association. 

“That progress reflects years of work by operators investing in new technologies, improving practices, and adapting to evolving requirements,” said Carly West, executive director of the trade group API Colorado.

Recognizing the success of the recent collaborative negotiations is important when Colorado lawmakers and regulators consider yet more rules, Granger added. They should “avoid adding costs or complexity that could hinder innovation, harm Colorado’s economy, or deliver little additional environmental benefit,” she said. 

Hickenlooper, now a Democratic U.S. senator, pointed to the Colorado rules under his governorship as a national model adopted by then-President Obama’s administration (and now being reversed by the Trump administration.)

“Colorado was the first state to enact smart methane rules,” Hickenlooper said. “The data proves they work, delivering cleaner air for our communities and making real progress in confronting the climate crisis.”

The EDF report said satellite data from a similar analysis from the Permian Basin also “provided strong evidence that New Mexico’s oil and gas methane pollution regulations were effective in reducing emissions” even while production there increased in 2024 and 2025. 

The Geo Tracker was a severely underrated runabout. I’ve driven a few of these before and have always adored their underdog spirit, their surprising off-road capability, and their oh-so-cute looks. Is it time for the Geo Tracker to come back? Oh yeah it is.

Thomas wrote about a new graphics package for the Toyota RAV4. TheDrunkenWrench got me really excited:

It is time. All the pieces are there:
-The people crave small, cheap, proper 4x4s
-The CAMI plant needs a model to save it’s production
-Vinyl graphics are back

THE GEO TRACKER MUST RISE FROM THE ASHES.

Matt wrote a Morning Dump that contains an interview with the legendary Bob Lutz, who held back no punches about modern BMW design. Ranwhenparked:

The headline could easily read “Ex-BMW Exec Bob Lutz Has Functioning Eyes, Brain”

Sid Bridge:

BMW Exec: Sir, I thought you should hear what Bob Lutz said.
BMW President: I hope this is worth interrupting my lunch.
BMW Exec: He said our cars are ugly.
BMW President: What does he know?! Now go fetch me some more onion rings to put on top of my Kimchi & Durian sandwich and don’t get any on the shag carpet. Oh, and check and see if my new “Best of Slim Whitman” CD came in. Say, why don’t you drop by this weekend and we’ll catch up on the the absolute best James Bond movies with George Lazenby. And go ahead and put a few bucks on the Dallas Cowboys for me. I think this is gonna be their year.

Jason wrote a Cold Start about the Peel P50, and I love how the advertisement shows what appears to be a woman meeting a guy for a coffee date by driving her car into the shop. UnseenCat:

Taking a P50 into the cafe for a date does make some sense. After all, if the datye starts to go badly, or he turns out to be a creep, she can just slam the door and peel right out …

Have a great evening, everyone!

Top graphic image: GM

The post The Geo Tracker’s Time Has Come Again: COTD appeared first on The Autopian.