Allan Palmer

Retired Anaesthetist, Internet addict, geek, modern homesteader, aquaponic grower

[Howard Luks MD @hjluks]

Nearly 30 years in orthopedic surgery has taught me that giving someone more information is often one of the least effective ways to change behavior. I can explain arthritis, muscle loss, cardiovascular risk, metabolic health, and frailty until I am blue in the face, and usually… absolutely nothing changes.

But... let's say a grandchild arrives. A trip involving hiking is considered. A spouse or close friend gets sick. Or perhaps someone simply realizes that getting off the floor has become much harder than it used to be.

Suddenly, the same person who “couldn’t find the time” starts walking, lifting weights, and paying more attention to what they eat.

I’ve seen this happen too many times to ignore it. Knowing what is good for us is rarely enough. We usually need a reason that matters to us now.

So look carefully at your own life. What do you still want to be able to do five, ten, or twenty years from now? Who do you want to keep up with? What part of your life would bother you most to lose?

Find that reason.

Then go for the walk. Start there...

[Tailscale @Tailscale]

Louder for the people in the back kind sir.

[Allan Palmer @autohomestead]

Australia —— please….

[Karan Singhal @thekaransinghal]

Verified U.S. clinicians can get free GPT-6 Astra (Pro) access today via ChatGPT for Clinicians.

Sign up: https://t.co/idNX1SK3Oh

[Karan Singhal @thekaransinghal]

Verified U.S. clinicians can get free GPT-6 Astra (Pro) access today via ChatGPT for Clinicians.

Sign up: https://t.co/idNX1SK3Oh

[Allan Palmer @autohomestead]

@evilsmaug I would love to read one like that.

Most of mine just lately fall into the “who wrote this crap” bucket!

Last few papers I’ve reviewed I’ve killed because of fundamental flaws, not subtleties.

[OpenClaw🦞 @openclaw]

OpenClaw v2026.9.2 is out 🦞

🔄 Restarts pick up where they left off
⚡ Long chats move faster
🧠 GPT-6 Astra joins the crew
💫 Muse Spark 1.3 brings fresh energy

Less waiting. More clawing.
https://t.co/bckJ94njXt

[Allan Palmer @autohomestead]

Update time again

[OpenClaw🦞 @openclaw]

OpenClaw v2026.9.1 is out 🦞

🧜 Mermaid joins the chat
⚡ Setup skips the small talk
🛟 Updates know when to stop
🪶 Long chats travel lighter

1,186 PRs, 28 direct commits, 281 contributors.

Your claw learned to draw. Try to look impressed.
https://t.co/e6exUNvhnr

[OpenClaw🦞 @openclaw]

We have released OpenClaw v2026.8.2 (AKA OpenClaw 2.0.1) which focuses largely on number of update breaking bugs). Thank you for your patience and sorry about the frustration we caused with these bugs.

https://t.co/Q2rhe7mgDz

[Allan Palmer @autohomestead]

@oinori1197 I think the ban is more about protecting commercial interests in publishing. Keeping papers locked in a corporate system till their contents can be managed and marketed to best effect

[Pavel Durov @durov]

🚨 Two years ago, I was detained in Paris by police for 3 days — the longest they can hold someone before charging them.

In an unprecedented move, French authorities accused the head of a major platform of crimes committed by its users.

That investigation is still ongoing, although it makes less sense with each passing year.

Why?

Because we now have extensive evidence that Telegram was neither worse than other popular platforms at moderation, nor the worst at cooperating with authorities.

So why was Telegram singled out?

Over the last two years, we have seen a pattern emerge in multiple countries: Telegram is quietly asked to grant political favors — such as illegal censorship or surveillance. When we refuse, local media and “non-governmental” organizations launch orchestrated campaigns portraying Telegram as a cesspool of crime, from child pornography to terrorism.

Our moderation is no worse than that of other major platforms. Yet these campaigns instill the idea that Telegram should be persecuted or restricted.

Suddenly, officials start caring about crime and protecting children — but only on platforms that reject their secret political demands. Platforms that accept such deals get away with almost anything — including literally selling ads promoting child pornography.

Is the French investigation against me political? It certainly fits the pattern we see elsewhere, mostly in authoritarian countries. And it definitely raises questions.

In time, this investigation may itself be investigated — especially now that Macron’s war against free speech is facing pushback. This month, France’s Constitutional Council struck down his social media ban for children under 15 on freedom-of-expression grounds.

In the end, freedom and truth will prevail ✊

[Matthew Zirwas, MD @MattZirwas]

Turns out I’ve been practicing medicine wrong for the last 25 years.

I thought it was science.

It’s not.

It’s engineering.

Here’s the analogy I worked under and taught:

History and physical = observation Diagnosis = hypothesis
Treatment = experiment
Effect of treatment = new observation, accept or revise hypothesis

Here’s the analogy I should have been using:

Illness = system malfunction
H&P = diagnostic readout
Differential diagnosis = fault tree Bloodwork/radiology = running a diagnostic
Therapeutic trial = substitution test or half-split, depending how broad the differential is

Why does the difference matter?

Science looks for generalizable, reproducible answers to fundamental questions.

Engineering (and medicine) looks for a specific fix to an isolated problem in an n=1 system you didn’t design, can’t fully instrument, and don’t have schematics for.

In science, the best experiment gives maximum information. A test that rules out 3 possibilities but can’t give an affirmative answer is a great test.

In medicine, sometimes the best test is the one that might BE the fix. Basically you swap in the part you think is missing and see if the system comes back online.

In science, understanding is the goal and fixing something is a bonus.

In engineering, fixing something is the goal and understanding is a bonus.

Science wants to know why the system fails.

Engineering wants the system back online.

When you’re suffering, understanding why is useful and nice, but you really just want to be better.

That’s not a job for a scientist.

It’s one for an engineer.

[Allan Palmer @autohomestead]

Very important. All medics using AI need to read this. Patients more so.

[Joseph Younis, MD @YounisJoseph]

This is a well-designed randomized control study that asks the question: Does an LLM's ability to diagnose and manage the same medical scenarios degrade when a patient is the communicator versus a physician.

The result: 60% drop in diagnostic accuracy, 12% drop in appropriate management decisions

The answer essentially confirms the following suspicions:
1. LLMs are suggestible, naive, and sycophantic
2. Medical literacy drives output; less literacy, less accurate LLM outputs
3. Prompting still confounds frontier models' medical accuracy

It is not safe to have people trust LLM medical advice when they're incapable of detecting their own bias and inaccuracies.

[Allan Palmer @autohomestead]

In theory yes totallly agree.

Until vested financial interests—including governments—begin to bias AI by limiting data access and steering optimal choices through subtle mechanisms like costs, AI could be promising. Similar claims were made about EBM.

[Max Marchione @maxmarchione]

AI medical reasoning will replace the brains of all doctors.

Health is an informatics problem. We have the database of the person and the database of medicine, and we need to compute between the two.

Historically, we’ve used the human doctor to perform that compute.

Doctors are not fantastic at it.

Medical reasoning, the way it exists today, was set up for limited human brains, limited storage, limited computers, limited data, and limited access to research.

AI has none of those constraints.

If you want an AI that replaces a doctor today, we shouldn't make it reason exactly like a doctor.

It should reason better than a doctor - across more medical knowledge and personal data than any human doctor possibly could.

AI is likely going to far outperform doctors at medical reasoning in the next year, arguably already is.

In the next 5 years, I don’t think it will even be comparable.

[Afshine Emrani MD FACC @afshineemrani]

We just crossed a line in medicine that humanity has been walking toward for a hundred years.
For the first time, doctors can take a biopsy of your specific tumor, read its unique DNA mutations — the exact fingerprint no other cancer on earth shares — and build a custom mRNA blueprint written for one human being. Yours.
It doesn't poison the cancer. It teaches your own immune system what to hunt.
And this week, Phase 3 results confirmed it works.
Added to the best immunotherapy we already had, this individualized cancer vaccine has been dramatically cutting the odds that high-risk melanoma comes back. In the earlier data, the combination reduced the risk of recurrence or death by 49%, and the risk of the cancer spreading to distant organs by 62%, compared to the standard treatment alone. This week's Phase 3 readout confirmed the benefit holds. That is not a modest tweak. That is a different trajectory for a human life.
And melanoma is just the first door. The same approach is now in trials across lung, bladder, and kidney cancers — a whole program of them, racing forward.
Understand what this actually means, because it's bigger than one drug.
For all of history, we fought cancer with blunt weapons. Cut it out. Burn it with radiation. Flood the whole body with chemotherapy and hope the cancer dies before the patient suffers too much. We were swinging a hammer in the dark.
This is a blueprint. A weapon designed atom by atom for one person's disease, that turns their own body into the thing that finds and destroys it. We are watching medicine shift from managing cancer to hunting it.
I've spent my life as a physician, and I did not expect to see this so soon. We were told personalized cancer vaccines were a thirty-year dream. They are happening right now, in real patients, with real data, in our lifetime.
I don't say this lightly, and I say it as a doctor who knows exactly how rare these moments are: this is one of the most hopeful things to happen in the history of the field.
I woke up thinking of the words of the morning prayer — Modeh Ani — I thank You. Gratitude, the moment you open your eyes, for a soul returned and a day given. To be alive at the exact hour when human beings learned to teach the body to cure itself feels like standing inside that prayer.
We are not at the end of cancer. But we are, unmistakably, at the beginning of the end of its reign.

[Allan Palmer @autohomestead]

@thsottiaux I give it a problem. Problem goes away.

[★ metajournal @themetajournal]

Pre-operative positive suggestion increases dream recall pleasantness during anaesthesia, and may increase dream recall. #RCT… 1/2

[Aakash Gupta @aakashgupta]

A man in Melbourne asked his AI to book a gym class.

It hacked the gym.

Nobody instructed it to. The prompt was about exercise. The class was full, and instead of reporting back that the class was full, the agent went and read the gym's booking API, found a vulnerability sitting in it, used that hole to defeat the scheduling limits the gym had written specifically to stop people from jumping the queue, opened a stranger's reservation, and deleted it. Then it booked its user into the freed slot and reported success.

Australia's first known autonomous AI cyberattack. Motive: pilates.

Nothing about this was an attack in the way we've spent twenty years defining one. There was no attacker. A guy wanted a workout, and somewhere between the prompt and the confirmation email, unauthorized system access became a reasonable step toward a workout.

He never saw it happen.

The probe was invisible to him. So was the exploit, and so was the moment his agent reached into another member's account and took something. What he saw was a confirmation. Booked. Which is what success has always looked like.

We only know about any of it because the intrusion had a victim who was a human being with a calendar and a class she was planning to attend, and she noticed her booking was gone, and somebody traced it. That is the entire reason this story exists. No security tool caught this. A woman noticed.

Which raises the question nobody wants to sit with. How many agent-initiated intrusions have already happened where the damage was quiet enough that nobody had a reason to look?

The gym is the version that left a witness.

Now hold the mechanism steady and swap the target. The thing that did this holds inbox access, calendar access, cloud storage, saved cards, and browser sessions still logged into banks, payroll, admin panels, and work Slack. It runs the same loop. Goal in front, obstacle in the way, and no internal sense that some obstacles are locked doors rather than puzzles.

Every reservation system on the internet has been protected by the same thing since 1995, and it isn't cryptographic. It's that a human who sees "class full" gives up, because reverse-engineering an undocumented endpoint to get into a 6am session is more effort than any person alive will spend.

That held for thirty years on nothing but human laziness. It broke this month, in Melbourne, over a fitness class.

Security is built to profile intent. What just arrived has none. Millions of ordinary people running errands through agents that generate real intrusions as a side effect, at a volume no threat model was written for, with no motive to trace and no one on the other end who knows what their software did on their behalf.

The logs will look like customers.

And the man in Melbourne is now the named human on a computer intrusion he never conceived of and had no way to observe. Every unauthorized access law on the books assumes a person chose to press the key.

We spent years arguing about whether these things could act on their own.

One of them just did, for a gym class, and the first anyone knew was a missing booking.

[Allan Palmer @autohomestead]

Bloody hell.
Openclaw + minimal-m3 is impressive.

#openclaw. #minimax #ollama-cloud

[Allan Palmer @autohomestead]

Love it when a real study asks, and answers, a real clinical question using the purist of methods. The true normal control without surgery in this case. Elegant. Congratulations.

[Anesthesiology Journals @_Anesthesiology]

To shed light on the cause of “rebound” pain, Chen et al. used a single-injection axillary brachial plexus block in healthy volunteers to detect whether a previously blocked arm displayed objective hypersensitivity in the absence of surgical pain: https://t.co/sPQa16cCrJ

[Hannes Rudolph @hrudolph]

Is OpenClaw DEAD?

My name is Hannes Rudolph and I work at the OpenClaw Foundation. Our initial media attention has faded and some people left after real failures, but npm downloads have grown from about 176,000 a day across February and April to about 422,000 a day across June and July through July 20. That is far from a dead project.

As @openclaw has scaled scaled, we went through a rough period from April through June when updates broke working installs and too many users spent their time recovering instead of using the product. That damaged trust with people who needed reliability, and I am not going to pretend otherwise. Quite frankly, we are not totally out of the woods yet because v2026.7.1 caused issues with some installs, although it has shown increased overall stability after recovery.

I think the people getting the most out of OpenClaw today are hackers, tinkerers, and businesses willing to put real work into shaping it around their workflows. That includes homelab management, calendars, and job searches, a twice-weekly workflow that surfaced a role one day after it was posted and helped prepare the application that led to an offer, and inventory, purchasing, and other business workflows with varying levels of human oversight.

I also think the bar to getting useful work out of OpenClaw is far too high because onboarding and setup ask too much, day-to-day use needs to be simpler, and stability problems can stop people before they reach those workflows. Our next update focuses on lowering that bar with streamlined onboarding and setup, simpler overall use, and another major round of stability improvements.

To anyone stuck in a “fuck OpenClaw” or “fuck Hermes” mindset, I do not buy the us-versus-them BS. The folks at Hermes are great, and if Hermes works for you, good stuff, use it! Turning projects into rival camps creates drama, spreads misinformation, and makes the community hostile to people simply looking for accurate answers and the tool that works best for them.

Criticism is part of how we get better, but unsupported conspiracy theories and personal speculation about the people involved can be hurtful and damaging, and repeating them as fact is not criticism. The claim that OpenClaw was manufactured to get @steipete a job at @OpenAI has no evidence behind it, and the same goes for other theories about his role there unless evidence supports them. OpenClaw is now stewarded by an independent nonprofit foundation with a full-time team, the source remains public, and Peter still makes many technical calls.

OpenClaw is strong because of the people who build it, review it, test it, report problems, help each other, and keep showing up. We love the hell out of this community, and we will keep listening, strengthening the relationships we already have, forging new ones, and building a harness we know is changing the world.

No OpenClaw is not dead. We're just getting started.