Open AI strikes back: ChatGPT's new feature will end AI slop

I have a source at OpenAI that leaked a major feature they're going to put out, that will make up the ground they're losing to Claude. Here's an early draft of the press release.

***

OpenAI is rolling out a new feature with ChatGPT model 5.5 that provides special resistance to an ongoing issue with AI output: the telltale signs that the output is AI, which is a major turnoff and source of frustration on the internet.

The new feature, dubbed Preview Plus(tm), lets users go through ChatGPT's output and actually purge all the AI pain points.

• Em-dashes? Users can delete them.
• "It's not X but Y"? It's "X in Y clothing"? Users can replace it with their own authentic version a human would actually say.
• Hallucinated legal citation? Users can insert something from actual case law that they have personally validated applies to the matter in question.

Preview Plus(tm) currently only available to paid plan subscribers and users who have opted in to early feature testing.

OpenAI CEO Sam Atman is hopeful about the new direction Preview Plus(tm) will take ChatGPT. Earlier today, he told a press conference, "Preview Plus empowers users to actually stand between ChatGPT and the eventual consumer of its output artifacts. It provides a seamless way to merge human intelligence with cutting-edge AI, getting the best of both worlds."

OpenAI gave a screenshot of an early version of the feature that shows a user removing the dreaded em-dash. 

 

 

The dubious lesson of the Carpathia

So right now I'm at LessOnline, the LessWrong conference centered on writing about rationality-related topics, and I encountered a very interesting proposition. There was a session about writing speeches for the LessWrong/rationalist Winter Solstice, a solemn ceremony with musical performances and speeches with a rational, hopeful-but-realistic theme. Towards the end, the speaker gave an example of a Tumblr post that he turned into a speech.

To summarize, it's about the sinking of the Titanic[1], and the captain of the Carpathia, who got the distress call and immediately went heads-down, focusing on the mission to get to the disaster as soon as possible, repurposing his entire ship and its resources for rescue, eventually saving some 700 lives. By pushing limits, he got there faster than should really have been possible, and thus rescuing far more than could have been expected.

A lot about this story rubbed me the wrong way, such that I don't endorse it for its intended purpose, which is to be an inspiring speech given at rationalist events in support of the rationalist worldmodel.

Look at this critical part of the exposition (italics original): 

He woke up all the engineers, all the stokers and firemen, diverted all that steam back into the engines, and asked his ship to go as fast as she possibly could. And when she’d done that, he asked her to go faster.

I need you to understand that you simply can’t push a ship very far past its top speed. ... Pushing a ship past its rated speed is not only reckless–it’s difficult to maneuver–but it puts an incredible amount of strain on the engines. ... They can’t do it. It can’t be done.

Carpathia’s absolute do-or-die, the-engines-can’t-take-this-forever top speed was fourteen knots. Dodging icebergs, in the dark and the cold, surrounded by mist, she sustained a speed of almost seventeen and a half.

...

They damn near broke the laws of physics, galloping north headlong into the dark...

Okay, something isn't adding up here. The post clearly emphasizes that ships are not supposed to be pushed past their top speed, and owns up to the fact that it really is reckless, and even acknowledges that the maintaining that speed might as well count as "physical law, violated". The reads as the fantasy of the manager who overcomes physical limits by sheer force of leadership ... which is not how physical systems, or tradeoffs thereon, actually work.

Also -- "dodging icebergs"? That's what got the Titanic into this mess in he first place. Maybe not the right image to use here.

All of that adds up to the fact that any success here really doesn't generalize. Knowing what they knew at the time, it was actually not the rational thing to do. Yes, it looks like a bold, selfless act now, because we know how it turned out: he happened not to sink his ship in the process. But he couldn't know that would be true at the time; for all he knew, this was condemning the ship and its crew to be casualties in addition to that of the Titanic (up to the rescue efforts of a later, more cautious ship). Claiming vindication reeks of hindsight bias. (There's a reason they caution rescuers against "being a hero" -- they may very well end up being yet one more person that needs rescue.)

One of the hurdles in making your thought more rational is accepting that something can be a bad idea despite the fact that it worked. In Blackjack, if you stand on 4, and the dealer busts, you win ... but it's still stupid. The idea that you should push a critical system well past its limits to accomplish a stretch goal is ... not good. Any success would, at most, be an edge case, not a canonical template you should keep repeating.

So ... if I were to make this a speech, it would have to be with some kind of twist like "lol gotcha! You just got seduced by a narrative, and hindsight bias. 'Push past the limits' makes for a rousing story, but in practice, it's how disasters happen." Which is the kind of chain-yanking I prefer not to do to begin with.

I'm not digging into the whether the presented facts are actually true, because I'm just evaluating it as a speech and working from assumptions it sets up. If there are other reasons that justify why this may have been less reckless than it really was, they should have been included to support the conclusion. But, as given, it pumps up the idea that, because you were bold, you had to save lives, and it works, that adds up to it being a good decision. It doesn't.

And there's a lesser issue, too. It also mentions repurposing part of the ship's steam to prepare tea and coffee for the survivors as they were loaded on board. Like, what? I'm sorry, but in a survival situation, coffee and tea are luxuries. They need warmth and drinking water ... the rest isn't a priority.

[1] Because I don't feel obligated to follow ridiculous, ossified conventions, I'm going to refer to ships as "it" and not italicize their names. They're proper nouns. Capitalization is good enough.

Looking too available is a security risk! The new NIST standard.

I've always been a little worried about people who post e.g. a Calendly link that shows their full availability and lets you set an appointment. Do you really want to reveal how not-busy you are? Well, apparently the NIST is now warning that it's also a big security risk, and advises a few protocols to obfuscate your true availability. I got my hands on an early draft of the proposal.

---

🛡️ NIST IR 8429-DRAFT

Presence Obfuscation in Federated Scheduling Systems

Guidelines for Temporal Metadata Minimization in Collaborative Availability Platforms

Issued by: National Institute of Standards and Technology
Prepared by: Information Security & Human Factors Division (IS-HFD)
Document Type: Interagency Report (Draft for Public Comment)
Release Date: April 1, 2025

1. Scope and Purpose

Modern calendaring tools often expose excessive availability to collaborators, contractors, and external partners. These systems—by default—permit observers to view large blocks of unscheduled time, often across multiple days, weeks, or recurring patterns.

While intended to facilitate meeting scheduling, such broad exposure of free/busy data introduces serious risks:

• Temporal profiling
• Passive inferences about workload or engagement
• Social graph deduction via availability overlap
• Identification of predictable solitude windows

These exposures disproportionately impact individuals with high meeting asymmetry—those who receive more requests than they initiate—and those whose roles rely on maintaining controlled perceptions of demand.

2. Problem Definition

The visibility of extensive availability windows has emerged as a key metadata leakage vector in both organizational and interpersonal contexts. Specifically:

• Open calendars reveal a subject’s unstructured time density, which may be misinterpreted as low workload or underutilization.
• Observers may make inferences about social graph positioning based on recurring exclusion from scheduled events.
• In adversarial contexts, such patterns can even aid in physical vulnerability modeling, including mapping of long-duration solitude intervals.

Note: While this report avoids normative language regarding “perceived busyness,” we recognize it as a functional privacy parameter in professional ecosystems.

3. Mitigation Strategy: Presence Obfuscation Layer (POL-1)

POL-1 is a modular framework for introducing structured ambiguity into exposed calendrical availability, ensuring that observed scheduling surfaces retain semantic plausibility without revealing raw temporal capacity.

It is not a deception system, but rather a context-aware redaction buffer for collaborative environments.

4. System Components

4.1 Observer-Coherent Subset Exposure (OCEP)

Availability is partitioned by observer class (e.g., internal peer, external client, HR metadata node) to ensure deterministic, plausible views that are not cross-correlatable. No single observer can reconstruct the true surface.

4.2 Continuity Entropy Injection (CEI)

Injects plausible placeholder constraints and pseudo-events to avoid the appearance of long unbroken availability, particularly during mid-day periods known to trigger unsolicited booking attempts.

4.3 Foreground Availability Perturbation System (FAPS)

Applies low-amplitude randomization at event boundaries to disrupt timing-based inference attacks. Designed to deflect automated meeting tools and high-frequency schedulers without impeding intentional collaboration.

Note: FAPS may be disabled in constrained or high-determinism scheduling environments (e.g., surgical coordination, launch windows).

4.4 Behavioral Availability Normalization Kernel (BANK)

Aligns exposed availability with industry-standard load models, e.g., “Product Manager (Mid-Level, West Coast)” or “Postdoctoral Researcher (Remote EU).” Useful for avoiding apparent availability outliers that may bias request behavior.

4.5 Fail-Safe Redaction Coherence Module (FRCM)

Final pass verifier that ensures no exposed schedule appears suspiciously empty, implausibly overbooked, or temporally inconsistent with the subject’s graph classification.

5. Application Scenarios

Scenario	Exposure Risk	Mitigation Stack
Early-career engineer in open scheduling org	Appears to have 4–5 open hours/day	BANK + CEI + OCEP
Public figure with assistants booking on their behalf	High-profile scheduling scrape risk	FAPS + FRCM
Remote worker in non-managerial role	Pattern of availability invites frequent “quick chats”	OCEP + CEI

6. Deployment and Compliance

POL-1 is suitable for deployment across major federated scheduling platforms (CalDAV, Exchange, Google Calendar API v3). Compatibility modules are provided for Outlook Graph Surfaces and GSuite Legacy Redirection endpoints.

Note: Organizations governed by EO 14217 (“Minimum Obfuscation Baselines for Federal Metadata Systems”) must deploy POL-1 or equivalent by Q3 FY25.

7. Availability

A reference implementation of POL-1 is available on the NIST GitHub mirror under a modified FIPS-aware BSD license:
📎 https://github.com/nist-opsec/pol1

"Availability is a feature. Apparent availability is a liability."
— NIST IR 8429-DRAFT, April 1, 2025

Hyperlinks for emails?

It's looking like we could be able to hyperlink to emails soon! No more of this "oh uh check your email for the one I sent at 11:27...". You'll be able to add a link that works just like linking the web!

The IETF is working on their "Parsable Mail Pointer Email Protocol" (PMPEP), and I was able to get an advance copy! Check it out below:

RFC 9869: Parsable Mail Pointer Email Protocol (PMPEP)

Personally, I think they backronymed it from "Per My Previous Email Protocol"...

Edit 4/11: Yes, this was a joke. And apparently it would indeed be enough just to use the messageID. So why hasn't anyone implemented this???

Setting short-selling straight; or, "But who let you short that?"

You might not be aware, but I've been short-selling some cryptocurrencies. (I would have said "been making money short-selling cryptocurrencies", but ...)

Often people ask some from of, "oh wow, what broker lets you do that?" It's actually an interesting misunderstanding, in that it misses a key insight:

You can short-sell any time you have a debt denominated in that asset.

At that point, you are short the asset. You benefit from anything that makes that asset easier to obtain, and thus extinguish your debt.

In the decentralized finance world, there are platforms (in my case, Compound.finance) that let you deposit some crypto asset A, and borrow some other crypto asset B. Once you do so, and sell B, you are now short-selling B!

The concept applies more generally too: for example, if you owe your friend the favor of helping them move, then you are "short moving services" (because anything that makes moving services cheap, also makes your debt easier to service, at least because you have the option to satisfying by paying a service rather than doing it yourself).

Also, if you borrow US dollars, and spend them, you are "shorting the dollar", although it's usually not talked about in these terms. (You hold a debt denominated in dollars, after having "sold" them for something else, which we generally refer to as "buying".) Although, there was an interesting case where people would borrow USDC (a crypto dollar substitute) and then find out they could be bought back for less than a dollar during the Silicon Valley Bank collapse. Thus making them "accidental short-sellers"!

A serious paper on bits as Joules per Kelvin

My ramblings about thermodynamics aren't so off-base, it turns out!

Remember this one? From 2009? Where I explained how Joules per Kelvin (energy per unit temperature) is a valid measure of information (or entropy, effectively "missing information"), which is normally measured in bits?

Well, now there's a paper that formalizes that idea and related ones. As the title ("Temperature as Joules per Bit") indicates, it looks at a rearranged version of the same insight (mine was "bits as Joules per temperature"). But, it also goes a lot deeper and derives thermodynamics starting from entropy to understand temperature, rather than the other way around, as is conventionally done.

Related thought: I remember back in that 2009 thermodyanmics/info theory frenzy, one of my goals was rederive the Carnot limit based on information-theoretic considerations -- that is, show it as a simple implication of the amount of knowledge you have about a system in a case where only know the temperature difference. (Naturally, I assumed someone had already done this and tried to find it but it was very hard to google for.)

Background: The Carnot limit tells you the maximum amount of mechanical work ("useful energy") you can extract from heat -- like, through an engine -- and, as it turns out, it's a function of the ratio of absoute temperatures you're working between. You don't face this limit when extracting work from a flywheel (spinning disc with grooves). Inspired by an counterintuitive insight in an Eliezer Yudkowsky LessWrong post, and my thoughts about it, I figured you could draw a more direct line from "knowledge of a temperature difference" to "how much energy is extractable".

Now I'll give it a go with ChatGPT, and post my findings!

ChatGPT4 refines my argument about fish being bad swimming instructors

I finally splurged for a paid ChatGPT subscription! With access to GPT-4, I asked it to refine my earlier argument about fish being bad swimming instructors. Here's what it gave me:

***

Certainly, let's rework the argument using more scientifically grounded explanations and references:

While it's true that fish, with their adept aquatic abilities, may seem like logical swimming instructors for humans at first glance, upon closer examination, it becomes clear that this is far from practical or effective. Even if we put aside the substantial communication and physical barriers, there are fundamental biological and behavioral differences between fish and humans that pose significant challenges.

Humans are terrestrial, bipedal mammals, and our motor skills and locomotive mechanisms are naturally adapted to this lifestyle. Fish, in contrast, are aquatic creatures designed to maneuver efficiently in water (Webb, P.W., 1982, "Locomotor patterns in the evolution of actinopterygian fishes"). Their ability to swim is an innate behavior, hardwired into their neural circuitry, not a learned skill (Gould, J.L., 1982, "Ethology: The Mechanisms and Evolution of Behavior"). Given this, fish cannot consciously comprehend or articulate the mechanics of their swimming abilities as humans would require in a teaching-learning scenario.

In the field of pedagogy, it is well accepted that the process of teaching involves more than simply demonstrating a skill; it requires an understanding of learning styles, the ability to give and receive feedback, and a degree of self-awareness in the teacher (Boud, D., et al., 1985, "Reflection: Turning Experience into Learning"). Fish, as far as current scientific consensus suggests, lack these cognitive capacities (Butler, A.B., et al., 2005, "Evolution of the Neural Basis of Consciousness: A Bird-Mammal Comparison").

Furthermore, many human skills, including swimming, are learned through a combination of instruction, practice, and feedback (Kirschner, P.A., et al., 2006, "Why Minimal Guidance During Instruction Does Not Work"). This model of learning doesn't align with the unconscious, reflex-driven behaviors seen in fish.

Lastly, it's important to acknowledge the vast physiological differences between humans and fish. Human limbs and fish fins have evolved for different purposes and environments, with humans developing more for grasping and manipulation and fish for efficient aquatic locomotion (Thorsen, D.H., et al., 2004, "Biomechanics of Locomotion in Asian Carp"). These differences are fundamental and cannot be bridged by simple observation or imitation.

In conclusion, despite their swimming prowess, fish cannot be effective swimming instructors for humans due to biological, behavioral, and cognitive differences, as well as practical communication and anatomical obstacles. This argument does not discount the value of observing aquatic creatures to improve our understanding of efficient movement in water, but it's clear that such observation is a far cry from active instruction.

Extended shower thought: Fish would be bad swimming instructors

Obligatory epistemic status: Speculative and hard to test but falls out an as implication under current worldmodel.

Just a shower thought I've had on my mind for a while:

Fish would be bad swimming instructors (for humans).

This remains true even after adjusting for/overcoming the differences in body types and ability to speak your language.

What do I mean? Well, to be sure, fish are good at swimming. And normally, that's what's you want in a swimming instructor...

...but not here. While fish are good at swimming, they have always been good at swimming. They were deft swimmers before they reached any point of being able to reflect on their selves or what they are doing. They have no appreciation for what it's like not to be a good swimmer.

What does "learning swimming" involve (for humans)?

Fish, being fish, never had to make the transition from being a natural biped to one that has repurposed its body for motion in water, which is what humans have to go through to "learn swimming". They never had to start from a mentality that finds walking somewhat natural -- and swimming somewhat unnatural -- and then adjusts its appendage motion in a way more suited to the latter.

We can imagine any kind of instruction session between fish and humans (as above, appropriately adjusted for language) to be fraught with peril. The fish might try to point out deficiencies that hold back the human. But it has never had to think about such deficiencies, because it has never needed to distinguish between the right way and the wrong way.  It was always just "the way" that came instinctually.

The fish might nearly lash out, frustrated that the human attempts an obviously ineffective means of water locomotion. But it has no idea why those flawed attempts are wrong. They just feel wrong.

The fish will happily demonstrate the "right way", just by doing it. But identifying the difference between its "right way" and "what humans do" would be a learning experience in itself, something it has no natural advantage in, as it was never part of the fish's "learning process for swimming".

The converse is true -- humans would be bad at teaching fish to walk (with the appropriate apparatus) or otherwise move on land. (But note the parallel isn't perfect as humans generally do have to struggle and learn how to walk, but not in any way that involves formal instruction.)

Implications for human-human interaction

So far, this insight feels (and, well, is) just idle speculation. But there are implications for ordinary life.

There will be times when someone is a "natural" at some skill. They're good at it. And that is their only teaching qualification. And they try to teach a non-natural that skill. They are then bad at teaching. The missing piece is the learner's mind is very different from what the natural is capable of filling in. Such a teacher will constantly show them how to do it "right" but not be able to identify the difference between what they're doing and what the learner is -- except by drawing on some kind of unrelated, general intelligence.

(I won't give any specific examples of such a skill, because those become contentious issues in their own right and detract from the general point. But I've definitely been on both sides -- learning from someone who has no understanding why they're good at it, and grasping to communicate something I do without ever thinking about it.)

You can also run into this problem when you become skilled at something. You can sometimes assimilate the skill so well that you are effectively a natural, by forgetting the whole process by which you learned it. You may lose the perspective you had as a beginner and are no longer able to relate to them.

Part of why I enjoy teaching what I know is that I seem unusually resistant to this process, and have vivid memories of the hurdles I overcame as a newbie.

Refuting arguments no one cares about: Exploiting the "help" in the pandemic vs the wildfires

I don't think anyone actually wanted to learn about this, but it's stayed on my mind for being "an argument that's wrong and I can prove".

Background:

A while ago I made fun of a Facebook friend (call her D) for her "let them eat cake" cluelessness. This was during the 2018 (?) California wildfires when she made a big post saying how thankful she was that, in miserable air conditions, she could "still get her groceries" through delivery services. That prompted me (and, among others, another FB friend, "E") to drop our jaws and say, "Um, you don't care that you're just offloading all that suffering to poor workers that can't afford to just stay home, and will be breathing the lung-destroying air in your stead?"

(And, to be sure, there's the argument that someone total utility is increasing by virtue of how said workers still have the option to expose themselves to risks for money, and the counterargument about "well why have OSHA that workers can't opt out of ..." which is its own topic but didn't really appeal to me or E at the time.)

So far, so good.

But later, during the pandemic, it came out that E, for similar reasons, didn't feel comfortable just getting delivery so she could stay home when we were being asked/mandated to.

Those ... situations don't seem analogous at all to me, and I don't think someone should feel bad for ordering delivery during a pandemic like in an air quality emergency like wildfires. Here's why:

A) In a wildfire, you are shifting all of the hazard of the smoke onto the people who bring your deliviers.

B) But in a pandemic, moving to delivery reduces the hazard for everyone, including the delivery people.

In Kantian terms: If "everyone did it", then everyone would still benefit in case B). But in A), all the avoidance by the rich "D" personas would be matched by losses to those who still have to deliver.

To elaborate on B: the way a delivery service works, every worker involved has less Covid-spreading contact than than if everyone were shopping at a grocery store. The warehouses that set up the goods for delivery can, for their part, refactor and apply inexpensive countermeasures to reduce those worker's exposure. Furthermore, with everyone moving to delivery, you get economies of scale, allowing everyone to afford the delivery service.

So, to me, it didn't didn't seem like you were doing anyone a favor out of solidarity to keep getting your grocerys through in-person shopping.

Leaked Google initiative: No more passwords!

I have an inside source that's claiming Google will be rolling out a new replacement for passwords and other secrets for authenticating users. They shared the upcoming blog post/press release with me. They're moving to a more "holistic" authentication system? Let's see if this pans out. In any case, here's the not-yet-released announcement.

***

Are you who you claim to be?

User logins protect websites from malicious actors, like spammers and trolls. So when you go online, only people with legitimate credentials can access the useful features of the site -- and others can't impersonate you. For years, you've used logins -- such as a username and password -- to prove to the site that you are who you claim to be, like this:

Some go even further and add a second factor to authenticate with, like an SMS code or one-time-password generator like you might have in the Google Authenticator app.

But, we figured it would be easier to just directly ask our users who they are -- so, we did! Following on our earlier success with No CAPTCHA reCAPTCHA, we’ve begun rolling out a new API that radically simplifies the login experience. We’re calling it "Credential-Free Authentication" and this is how it looks:

On websites using this new API, a significant number of users will be able to securely and easily verify their identities without (separately) having to provide credentials: no password, no rotating code. Instead, with just a single click, they’ll validate who they claim to be.

A brief history of user authentication

While the new login API may sound simple, there is a high degree of sophistication behind that modest interface. Authentication has long relied on attackers not having critical secrets, like a password or random number generator seed or other private information. You may have heard the traditional formulation, that authentication requires you to provide something you have, something you are, or something you know.

However, our research recently showed that it's about as likely for the genuine user to be missing the credentials as it is for a malicious actor. How many times have you forgotten your password or encountered a bug with your password manager? (Not GPM, of course!) Thus, challenging users for credentials is no longer a dependable test.

Furthermore, attackers are often able to steal user credentials, forcing providers to rely on a secondary layer of fraud identification, so as to lock accounts when users behave suspiciously. You've seen this if you've ever had a credit card declined for an unusually large or remote purchase.

Introducing Credential-Free Auhentication

That got our security engineers thinking: if we already have to analyze a user's behavior in order to catch account compromises, why not just use that as the authentication? It would cut two carrots with one knife! After all, an attacker might be able to guess your password or your credit card information, but they will never be able to mimic the full depth and breadth of how you interact with websites, from your browing history, to your cookies set, to the way you move your mouse.

Following the "No CAPTCHA" model above, we developed an Advanced User Analysis backend for logins that actively considers a user’s entire engagement with the the Internet to determine who that user is. This enables us to rely less on "Do you have the secret?" and, in turn, offer a better experience for users. Now, users can just click a radio button, and in most cases, they’re logged in. In fact, you'll rarely have to log in at all, because sites will "recognize" you, just like you don't have to show your ID to go into an event venue a second time if the bouncer recognizes you.

But are you really that person?

However, authentication challenges aren't going away just yet. In cases where our tracking cookies and other behavioral metrics can't confidently predict who someone is, we will prompt the user for additional information, increasing the number of security checkpoints to confirm who the user really is. For example, you might need to turn on your webcam or upload your operating system's recent logs to give a fuller picture.

Adopting the new API on your site

As more websites adopt the new API, more people will see Credential-Free Authentication. Early adopters, like Snapchat, WordPress, Twitch, and several others are already seeing great results with this new API. For example, in the last week, the number of support tickets for account resets on WordPress went down by 90%. Twitch reported similar figures -- and also was able to unmask several sockpuppets who had been manipulating discussions and vote totals.

To adopt the new CFA API for your website, visit our landing page for more.

Good users, we'll continue to keep the internet safe and easy to use. Bad users, it'll only get harder to hide yourselves and take over legitimate accounts -- sorry we're (still) not sorry.

***

Edit: Yes, this was an April Fools joke.

How they handle multi-episode stories in a TV series

Related to: Expecting Short Inferential Distances (yes, linking LW is still a thing).

Below are some remarks I've made in scattered form in several forums, but I thought I'd collect them here.

Problem: A TV series will run for many episodes, and the writers will want to build up a storyline over many of them, which is necessary for a satisfying payoff. But viewers don't necessarily watch it all at once and keep the whole thing in their heads.

I call this the "narrative equivalent" of expecting short inferential distances (see the link above) -- just as people are used to an explanation not requiring a lot of steps, they don't expect a given scene or episode to need a lot of previous viewing to understand.

So, how do writers solve this problem? Here are the four general ways, with examples (feel free to offer more TV series for a category!):

A) Don't bother #1: Each episode is self-contained.

This is known as the "episodic" approach, as opposed to serial. Each episode can be understood without knowing anything about the preceding episodes, so there's no need to worry about this problem at all.

Examples: Star Trek (at least the original or TNG), South Park, The Simpsons (earlier seasons)

Downside: It's hard to feel investment when you know nothing will matter, that things will just return to where they were at the end. It also limits how much build-up (and corresponding payoff) the story can have.

B) Don't bother #2: It's the viewers' job to keep up.

Each episode just assumes you have the entire previous history in your head, maybe with some small reminders of previous relevancies for assistance.

Example(s): Game of Thrones

Downside: It only works for really devoted fans who will binge the whole thing and try on their own to stay up to speed.

C) Formal recaps

You've seen them: the narrator says, "Previously, on [this TV series]...", and then you get enough short clips to establish the relevant context for current episode.

Examples: 24 (after season 1), Battlestar Galactica, Burn Notice

Downside: A lot of people don't like them and think they're cheesy. (I've never understood this mentality, but there it is.) It also may force you to reveal what things from previous episodes will be relevant, taking away their surprise when revealed. There is a slight break in immersion since you have to go "out of the world" for them.

D) In-world recaps

Same as the previous, except you're not taken "out of the world" to do it; instead, there is a scene, within the story, that doubles as exposition of the things a formal recap would cover.

Examples: Breaking Bad, Bojack Horseman

Downside: It heavily constrains how you write the story and forces in pointless scenes that shatter immersion because they're retelling things -- and more slowly! -- the characters should know.

Well, there you have it. That about summarizes the different ways writers handle (or avoid handling) long storylines!

An HTTP status code to say "you messed up but I'll handle your request anyway"

So apparently, the Internet Engineering Task Force is going to introduce a new HTTP status code. Just like there's the 404 for "File not found", we're soon going to have "397 Tolerating", similar to a redirect.

The way it would work is, if you send a request that violates some standard, but the server can identify the probable intent of your request, it will reply with a "397 Tolerating" to say, "oh, you messed up, and here's how, but I'm going to reply to what I think you meant".

This is much better than the options we had before, which were either a) unnecessarily reject the request, or b) silently reply to the intended meaning but with no notice that was happening. This lets you tell the client you're tolerating their garbage!

My contact at IETF send me an early draft of the RFC, which you can access at the link below.

RFC 8969: HTTP Status Code 397: Tolerating

Pastebin Link

My presentation on using SAT solvers for constraint and optimization problems

Because of the virus we had to hold this meetup virtually, and I was slated to present there for the Evening of Python Coding. Since we made a recording of it, I can now share. Enjoy my not-ready-for-prime-time voice! (Yes, I need to update my profile picture ... badly.

Solving my first (Ghidra) reverse engineering challenge

I was pointed to this challenge by this article, and had heard about the NSA's new Ghidra reverse engineering tool. I was able to solve it without reading much of the article! Here's what I did.

Setup: They give you a compiled ELF binary that runs a program that prompts you for a password. It will tell you if you guessed correctly.

I was going to use this as a change to learn the Ghidra tool, with help from the article.

First problem: It's compiled to run on Linux Debian x86. I was doing it on a MacBook. So, it wouldn't run.

I noticed that Ghidra offers you the option to export the binary as C code (which makes sense, as part of Ghidra's reverse engineering is to convert a binary into assembly and slightly-more-readable C-like code). So, I figured I could just compile that to run on my Mac.

It didn't work though, since Ghidra exports an incomplete version of the code that uses some invalid types, and doesn't define all of its labels properly, which required a lot of manual work that was increasingly appearing like it would take too long.

So I figured I'd just run the binary in a Docker container. I found the Debian i386 version and pulled it down. (I had a long side diversion here getting a setup so I could edit files on my machine that would appear as I want in the Docker container, but the details are uninteresting besides this clever hack for getting the "copy file" command into your docker one-liner.)

So, I was able to run the binary.

Second problem: Somehow, it thought I was using a "debuguer" -- which I would be, at some point, but is strange, because I wasn't using one yet:

Don't use a debuguer !

I looked through the decompiled code to see where it might be doing this and found:

lVar1 = ptrace(PTRACE_TRACEME,0,1,0);
if (lVar1 < 0) {
    puts("Don\'t use a debuguer !");
    /* WARNING: Subroutine does not return */
  abort();
} 

Hm, okay, well that "if" statement maps to this part of the assembly:

0804868c 79 11 JNS LAB_0804869f

That is, according to this handy reference, "Jump if not sign." Well, whatever "sign" is, I want it to do the opposite -- change "JNS" to "JS", so it "jumps if sign" and skips that block -- the one that gives the mean message and exits the program.

I haven't figured out a good way to format that line, but 0804868c is the (hexadecimal) location within the binary, 79 11 is the hex version of the binary command itself, JNS the assembly term (just a mnemonic device) for that command, and LAB_0804869f is the argument passed to the command -- in this case, a label for where to jump to in the program.

According to that spec, the JNS command maps to 79, and if I want it to be JS, I need to replace it with 78. But...

Third problem: I don't have an easy way to edit binaries.

A convenient way to look at them is as a hex(adecimal) dump in a "hex editor", but I hadn't done that for a few months. Hex editors are useful because they show you the raw hex, the offset where it appears in the file, and, off to the side, the ASCII/text equivalent of that hex. Fortunately, I found this comment, showing how to repurpose my text editor, vim, to double as an easy hex editor. I can just open it up, edit the hex values, save, and it updates everything.

You can then run a utility, "gobjdump", to see the code as assembly and verify that e.g. JNS changed to JS as expected.

Fortunately, when I ran that edited binary, it did what I wanted: not accuse me of using a debugger, and proceed with the rest of the program.

The rest of the hurdles are similar: there's some if-test that can send the execution into a block that you don't want it to. Fortunately, this binary is structured ... stupidly, from a security perspective. It has that same kind of if-block for validating whether you entered the right password. Right afterward, if the check succeeded, it prints the password. Not your input, no -- it prints the correct password, which you can later validate on an unmodified run.

So, it's just a matter of tweaking the code so that the you always enter the block that outputs the password. (They were smart enough not to have the password itself appear as an obvious string in the binary, at least.)

Once I saw the output, I could submit it at the challenge site and very it was correct.

Now for something harder!

NAND to Tetris: Nothing short of awesome

Been a while since I posted, huh?

Well, here's an interesting development in my life: I finally started the NAND to Tetris course (part 2) on Coursera, which is based on the instructors' site.

I can say, whole-heartedly, that the course is awesome, and the first thing in a long time that has gotten me into a flow state.

The idea behind the course is this: they take you through (virtually) building a computer from a very basic level, showing all the abstraction layers that produce a computer's behavior.

It starts with logic gates -- specifically, the NAND gate (hence the name), since it's universal. Your first project is to use a hardware simulator to build several other logical gates from NAND. (NAND is just an AND gate with the outputs inverted, so that true and true yield false, everything else yields true.) The second project is to to build the arithmetic logic unit (ALU) in a CPU out of the components, so you basically have a configurable circuit: based on six control bits, you perform one of several possible functions on two 16-bit inputs.

The next project incorporates "flip flops", which allow you to repeatedly execute that circuit while also writing to and reading from some some persisted memory. Your inputs to the circuit then function as a kind of machine code.

The project after that then has you implement functions in that machine code, writing in an assembly language the authors created for the course that has a precise mapping to the machine code inputs in the CPU from the previous lesson. I can honestly say it was a really fun project to implement multiplication directly in assembly language!

Later on you write a compiler from a high level language into assembly (which then converts simply into machine code), in a way that's broken into two steps: a compiler from the high level language into a virtual machine that works on a stack with memory blocks, and a compiler from the virtual machine commands to assembly. I just recently finished that latter part (which comes first).

Those layers then build up to making a game that runs on your CPU, then an OS, and some other stuff I haven't delved into.

In the course of the projects, you use a hardware simulator, an assembler (for converting the assembly language into machine code), a CPU simulator, and a virtual machine simulator.

In order to catch up with the class an have enough leeway for a weekend trip, I went through much of the course over the weekend, and enjoyed every minute of it! I especially like how the break the projects into manageable pieces. For example, in the VM-to-assembly part, it first has you implement simple push/pop/add operations on a stack and run a test to verify that you can compile those commands. Further test suites give you a manageable set of operations to add.

I've written an actual compiler now! (albeit limited use...)

(Consider how fun this is, and how naturally it comes to me, maybe I picked the wrong field.)

One interesting challenge is that the CPU only has two registers call A and D, and only the A register is used when accessing memory, to know which word of memory to look at. It took me a while to figure out how you could say "look at the value in the memory location refered to in the memory location zero." Before I saw how you could do it, I implemented one project by having two separate code branches for the two possible values at the memory location!

Would love to link my code for the projects, but they discourage that to leave the challenge for others to solve.

Another Slashdot memory: Ah, the brick-and-mortar analogies...

So, I remembered another funny Slashdot exchange (again, no link).

Story: Some online retailer got in trouble for filtering their "customer reviews" so that only the positive remarks (about listed products) stayed and everything else was deleted.

A: "Wow, that's pretty scummy. They don't have the right to just clip out negative reviews."

B: "Don't they? I mean, it's their site; they have the right to set whatever editorial standards they want."

C: "Sure, but there's still an issue of consumer fraud and deception. For a brick-and-mortar analogy, imagine that Barnes and Noble started hosing 'book discussion nights' at their stores and promoted it as such. But you quickly notice that whenever someone says something negative about a book sold by B&N -- and only those books -- that person gets a tap on the shoulder from security, pulled aside, and asked to leave.

"In that case, it would indeed be correct to say they can expel whoever they want, but it's still fundamentally fraudulent to represent that event as being for 'book discussion' rather than 'book promotion'."

---

In other news, I finally found one of the ones that I thought I couldn't! The Armadillo rocket failure mentioned in this post was actually this conversation. The actual (but truncated) exchange went like this (note the links to original comments):

A: "And to think, they want us all to ride in these things commercially...."

B: "John and his team have an excellent track record thus far, and have continued to make safety a main issue. I'm sure that this experience will teach them even more, helping to make the next flight even safer."

C: "You mean even safer than a huge orange fireball?

"I don't know, that's a pretty high bar."

Some of my geeky tech jokes -- with explanations!

I know the line: explaining a joke is like dissecting a frog; you understand it better, but it dies. Still, not everyone will get these, and I figure I might as well have a place where you at least get a chance. So here are some of my own creations, explained.

---

Girl, you make me feel like a fraudulent prover in a stochastic interactive zero-knowledge proof protocol ... because I really wish I had access to your random private bits!

Explanation: In a stochastic zero-knowledge proof protocol, there is a prover and a verifier, where the former wants to convince the latter of something. But for proof to work, the verifier must give the prover unpredictable challenges. Think of it like a quiz in school -- it's not much of a quiz if you know the exact questions that will be on it.

The information to predict the challenges is known as the verifier's private random bits Those with a legit proof don't need this, but a fraudulent prover does. Thus, a fraudulent prover in a stochastic interactive zero-knolwedge proof protocol wants access to the verifier's "random private bits".

---

A historian, a geologist, and a cryptographer are searching for buried treasure. The historian brings expertise on practices used by treasure hiders, the geologist brings expertise on ideal digging places, and the cryptographer brings expertise on hidden messages.

Shortly after they start working together, the cryptographer announces, "I've found it!!"

The others are delighted: 'Where is it?'

The cryptographer says, "It's underground."

'Okay, but where underground?'

"It's somewhere underground!"

'But where specifically?'

"I don't know, but I know it's underground!"

'Slow down there. If all you know is that it's underground, then in what sense did you "find" anything? We're scarcely better off than when we started!'

"Give me a break! I just gave you an efficiently-computable distinguishing attack that separates the location of the treasure from the output of a random oracle. What more could you want?"

Explanation: In cryptography, an encryption scheme is considered broken if an attacker can find some pattern to the encrypted message -- i.e. they can identify telltale signs that it wasn't generated by a perfect random number generator, a "random oracle". Such a flaw would be called a "distinguishing attack". So in the cryptography world, they don't care if the attack actually allows you to decrypt the message; they stop as soon as they find non-randomness to the encrypted data. Applied to a treasure hunt, this means they would give up as soon as they conclude that the treasure location is non-random, which the cryptographer here things s/he's done simply by concluding that it's "underground".

---

So, 16-year-old Johnnie walked into an Amazon Web Services-run bar...

"Welcome," said the bartender. "What are you drinking?"

Johnnie replied, 'What've you got?'

"Well, we have a selection of wines and the beers you see right here on tap. But if you prefer, we also have club soda and some juices."

Johnnie thought, Wait a second. Why is he telling me about the wines and beers? Does he even realize ... ?

'Okay, I'll take the Guinness.'

"Bottle or draft?"

'Draft.'

"Alright, and how will you be paying?"

Johnnie only had large bills from his summer job and gave the bartender a C-note.

"Sorry, but I gotta check to make sure this is real." The bartender took out a pen and marked it, then counted out the change. Johnnie reached for the beer.

"Hold on a second! Make sure to use a coaster!" The bartender slipped one under the glass. "Okay, now enjoy!"

Johnnie lifted up the glass to drink. Before he was able to sip, the bartender swatted it out of his hand.

"WHAT ARE YOU THINKING!?! Don't you know 16-year-olds can't drink!"

Explanation: On the AWS site, they will gladly let you click on the "Launch server" button and go through numerous screens and last-minute checks to configure it, and only at the very last stage does it say, "oops, turns out you don't have permission to do that" -- so it's like a bartender that takes you through a entire transaction, even verifying irrelevant things (like whether the money is real), while knowing the whole time he can't sell to you.

---

How is a Mongo replica set like an Iowa voter?

In primary elections, they only vote for candidates they think are electable!

Explanation: Databases can have "replica sets" where there are multiple servers that try to have the same data; secondary servers depend on an agreed-upon "primary" to be the "real" source of data. Often times, the primary server goes down, so they have to decide on a new primary, known as a "primary election". But there are some restrictions on who they will vote for -- if they e.g. have reason to believe that a server can't be seen by other members, and in those cases it will regard that server as unelectable. So you can get funny messages about "server42 won't vote for server45 in primary election because it doesn't think it's electable".

More funny and insightful Slashdot posts I remember

If you liked the last post on this here are some more you might like. This time, I think they're more insightful than funny, but often times, insight is funny!

Also, I thought I'd point out the value of forums: in all of these exchanges, it seems to take three people to get the "aha" moment, not just one or two.

---

Topic: some new superstrong carbon nanotube material is discovered.

A: So it seems like they're planning to use this stuff for armor. But what about weapons? It seems that anything strong enough to make good armor would also have value as a weapon.

B: That doesn't follow at all! They're completely different use-cases. For example, leather is known to make good armor, but you never see a leather sword!

C: Sure you do -- it's called a "whip" and we dig them up all the time!

---

Topic: Some dating site is blocked from emailing University of Texas students (at their school email) because of too much spam.

A: Well, their defense is that they're complying with the CAN-SPAM Act, in that they have an unsubscribe link, use their real name, etc., so UT can't just block them wholesale.

B: What? That doesn't matter; that's just the *minimum* requirement for sending such emails. Obviously, any domain owner can impose whatever extra restrictions they want!

C: Yeah, it's like saying, "I have a valid driver's license. I am the legal owner of this vehicle. I have paid the appropriate taxes and registered it. I hold the required liability insurance, and the vehicle is in good working order. I have complied with all applicable traffic laws and operated it in a safe manner. Therefore, I have the right to park on your lawn."

---

I actually use C's last example when explaining to people the difference between authentication (proving who you are) and authorization (what that identity is allowed to do).

If the minimum wage prohibitions are so easily circumvented ...

The recent Talia Jane story just made me realize we have a possible inconsistently in policy. To get you up to speed, Jane took a low-wage job in the San Francisco Bay Area, hoping to work her way up to her passion of being a social media manager for a major company. But because of rental prices, she paid 85% of per post-tax pay just for rent (!), complained about her employer paying so little, and then was fired.

But as for the inconsistency:

Illegal: paying someone below $X/hour.

Legal: paying someone ($X + $Y)/hour (Y positive) to work in a place where their discretionary income would place them in extreme poverty (e.g. 85% of post-tax on rent).

And yes, that's just an (arguably trivial) corollary of "minimum wage (and tax brackets for that matter) is not automatically cost-of-living-adjusted". But if the goal is to stop people from being taken advantage of with low job offers that hold them in poverty, that seems like a pretty big loophole.

And it's not just that -- let's say someone moves farther out to be able to afford to live there. Then they're traveling an extra N hours just to make each shift which should rightly count against their effective hourly wage.

So, food for thought: what are we really trying to optimize for here? What would the law have to look like to not just avoid these loopholes, but "carve reality at the joints" such that it's fundamentally impossible to scalably circumvent such a law?

If you keep raising the minimum wage for a locality, and people keep commuting greater distances to get that income, what have you accomplished?

Funny Slashdot exchanges, before they're lost to time

In the time that I was a regular reader of Slashdot, I saw a few exchanges that stayed in my mind. I later went back to find them, but was never able to. So that they're not lost to time, I figured I'd post all the ones I remember. What follows is from memory, and prettied up a bit. (Not trying to plagiarize, if you can find the original post for any of these, let me know.)

Enjoy.

---

[Story: Armadillo Aerospace has a failed rocket launch.]

A: Well, I think we can close the books on Carmack's little project.
B: Come on, now. Private space travel is still in its infancy. There are growing pains. Not everything works the first time. But what's important, is that we're learning from these events. Armadillo is learning. They'll adapt. And the next voyage will be better and safer!
C: You mean, even safer than a big orange fireball?

---

A: [long rant] So that's the problem with this ban on incandescent light bulbs.
B: Whoa whoa whoa, slow down. There is no "ban" on incandescent light bulbs. It's just that the government passed new efficiency standards, and incandescents don't meet them.
C: Oh, that's clever! I should try that some time: "See, I'm not breaking up with you! I'm just raising my standards to the point where you no longer qualify."

---

[Story: a pedophile was caught because he took pictures of his acts and tried to blur out the victims' faces, but police analysts were able to unblur them.]

A: Hah! What an amateur! Everyone knows you have to do a true Gaussian blur to destroy the information content of the picture!
B: Yeah, or entropize it by blacking out the whole face.
C: Right. Or, you know, you could just ... not molest children.

(IIRC, C was heavily voted down and criticized for assuming guilt.)

---

[Story: police used "big data" analytics techniques and discovered that most robberies occur on paydays near check-cashing places, which allowed them to ramp up arrests.]

A: I don't know, this seems kind of big-brothery...
B: Not at all! This is the kind of police work we should applaud! Working only off publicly available, non-private data, they found real, actionable correlations. It wasn't just some bigoted cop working off his gut: "Oh, this must be where the thugs go ..." No, they based it on real data. What's more, it let them avoid the trap of guessing the wrong paydays, which can actually vary! Some people get paid weekly, some biweekly, some of the 1st and 15th. For example, I get paid on the 7th and 21st.
C: So, uh ... where do you cash your checks, by chance?