> Ethical disclosure is a complicated topic, because researchers shouldn't hold bugs for ransom or demand high payment.
Why not? Capitalism requires they maximize their value. These profitable companies lay bare at the altar, so they should understand the requirements of their god.
I wouldn’t be surprised if you did and still had the same issue, but if you relied on the training data recall alone, you definitely shortchanged yourself.
I’ve had decent results with tasking a model to run pre-research and summarization as a checklist, then have a second instance of the model(s) review and then develop my learning plan learn, versus the times I just asked Claude or ChatGPT to explain something to me “from memory.”
That’s illuminating. I was born and have lived in the U.S. my entire life. If you asked me what the first day of the week was, I would immediately respond with “Monday”.
I’m going to mull this one over a bit more to figure out where my social education went wrong!
> The end of the end of the week cannot be the beginning of the week.
Of course it can. The word “weekend” just refers to the block of free time starting at the end of the working week. It doesn’t have any direct relation to the way weekdays are laid out on a calendar. The word was invented in England at a time when the English still considered Sunday the first day of the week, a hundred years before ISO 8601 moved Monday to the start for commercial reasons (as described in Annex A).
When a public holiday falls on Monday, my friends and I call that weekend a “three‐day weekend” or a “long weekend,” even though the Monday falls well within what a strict ISO 8601er would consider the next week. What makes the weekend the weekend is when the weekend starts—the end of the work week—not when the weekend ends.
Dictionary.com, Merriam-Webster, and Cambridge all define the weekend as some derivation of "the end of the week".
I'm also not sure how the various definitions and historical context has anything to do with the claim that the end cannot be the beginning. If the end were the beginning that particular day would belong to two different weeks, but that's nonsense. The last day of the prior week and the first day of the current week are two different days, therefore:
> The end of the end of the week cannot be the beginning of the week.
The weekend has always been innately tied to the work week specifically, both historically and in current usage.
> Dictionary.com
Has as the third listed definition: “any two-day period taken or given regularly as a weekly rest period from one's work.” And has a whole paragraph describing usage, including: “More strictly speaking, the weekend is thought to consist of Saturday and Sunday (often regardless of whether the calendar week is considered to begin on Sunday or Monday). In practical terms, the weekend is typically considered to be the period between the end of the workweek (or school week) and the beginning of a new one—which is why most people consider their weekend to start on Friday night, after work or school.”
> Merriam-Webster
“the end of the week; the period between the close of one work or school week and the start of the next”
> Cambridge
“Saturday and Sunday, or Friday evening until Sunday night; the part of the week in which many people living in the West do not go to work”
Every dictionary you cited agrees that the weekend is or can be defined in relation to the work week!
> If the end were the beginning that particular day would belong to two different weeks, but that's nonsense.
Millions of people living in cultures where Sunday is the first day of the week have no problem with the concept of a single weekend consisting of a Saturday from one week and a Sunday from the next. There’s no logical contradiction. “Monday as the beginning of the week is the only logically defensible position if you support the weekend” is a false statement, because a single weekend is not required to consist solely of days from the same week.
Officially (or historically, or to banks maybe) it starts Sunday, but to anyone I have ever talked to in the US its definitely Monday. Saturday and Sunday are the weekEND after all.
Wasn’t that the entire premise of the product when it launched? Disappearing messages and minimal footprint so you could be confident your communications were semi-private?
That was the expectation of using the product so it fits and isn’t anything like this discussion.
Yes. And because of the platforms deciding to enable that business model, we're now in a bullshit reality where the very apps that need to be screenshotted most often, are first to abuse platform features and prevent screenshots from being taken.
This deserves a longer rant, but the very premise of Snapchat was always poisonous, and is largely responsible for why, 20 years later, we're now facing extreme proposals for regulating electronic communication.
I generally disagree with the whole line of criticism techies get for "solving social problems with technology", but if there's one unambiguous case where I'd agree it was plain stupid, it's the concept of screenshot prevention, that became prominent with Snapchat. Controlling whether your recipient gets to keep the message you sent them is a purely social problem, and the answer to that in the real world has always, for good reasons, been "once sent, it's no longer yours; once received, it belongs to recipient to do with as they please".
> but the very premise of Snapchat was always poisonous
Wdym? Hooking (primarily underage) people onto "daily streaks" of sharing (often inappropriate) pictures of themselves to (sometimes random) other people via likely insecure servers is a _great_ business model!
Great customer retention. Lots and lots of engagement. So much shareholder value.
Oh and just think about how vulnerable they are. _Perfect_ for ads.
Their entire color story was really cool at the time. Yellow with Magenta/Blue accents was super bold, especially compared to the boring ass direction everything was going. Material and Windows 10 flatshit was all the rage and here comes Snapchat like the Kool-Aid man.
I can't speak for the OP, but in my research, the translation works out to be:
"Let me die with the foreigners"
I usually see "foreigners" replaced with "Philistines" (as noted above). It's Samson's line just before he braces against the pillars and brings the building down on himself and the thousands of Philistines inside, so I understand why it is usually translated that way.
Also, ψυχή here renders Hebrew nephesh, which is closer to "life" or "self", but when I think about the attributed statement it just reads better with the personal attachment (vs. "Let my self|life|soul die with..."). Which is the fun of translation, as each translator seeks to reveal a meaningful rendition which will resonate with their audience.
I don't think that disagrees with what I said. I was translating the Greek, not the Hebrew. But you're right the Hebrew פְּלִשְׁתִּים֒ is unambiguously "Philistines."
The Greek quoted above is from the Septuagint, and those translators rendered that word as ἀλλόφυλοι, roughly "those of another tribe." So "foreigners" is the closest I have in English to the Greek word, and "Philistines" is who it refers to.
I'm no scholar here, just a dabbler, so I'm happy the discussion is happening at all.
It's funny to be randomly reminded of the phyles from The Diamond Age, also from φυλή (it's an English word borrowed directly from the Greek, but not a very familiar one in many contexts today).
I notice the models with reasoning can conflate “internal” (or subagent) discussions with external (i.e. me). So it is accurately indicating “I’ve had this discussion before” but incorrectly asserting who it was with.
My understanding of how “thinking”works is limited though, and given the reduced visibility into the thinking traces, it is harder to tell if this is actually happening or if these are imaginary discussions the model for some reason calcifies on.
Oh, that's interesting - because that's absolutely what's happening in my experience.
If I look at the thinking (which seems to have become unavailable in Opus 5 a lot of the time, but was present - and often useful - in 4.8/4.6) you're right - it's having the discussion with itself, and seems unable to distinguish that discussion from discussions with me. BUT it also seems to be related to the length of the chat - this seems far more likely to happen in a longer chat.
I don't understand why they have removed visibility into thinking - I found it very useful, not only for spotting things like this, but also because in more complex discussions it would often mention (useful) things in its train of thought that it dropped from its response - but if I said "when you were thinking, you mentioned this" it would then expand on that point. Taking that away is another thing that has negatively impacted the value I get from Opus 5.0 versus earlier models.
With GPT 5.6 Luna the thinking once or twice leaked into the output for me. It's interesting, but perhaps not particularly useful.
It would be endless paragraphs of something among the lines of:
Need prepare final response? Yes provide. But wait, chat tool complete? Final needed but user already complete. Need summary, preparing final. Response complete. Wait but is final response complete? Need provide. Start finalizing now but wait did user acknowledge final complete? Assistant response final: user complete. Should now create final?
The reasons thinking traces are pretty much gone is, supposedly, to prevent distillation. Whether that is actually true or just an excuse is up in the air (because I at least am not going to trust Anthropics claims on why they do it).
And because the thinking, which does makes it better at achieving outcomes, contains naughty words and information (e.g. private information). https://stolen-thoughts.com/
"Thinking" is just normal model output that's hidden from user. In practice it's just stuff in a <reasoning> tag or similar that gets filtered out from the user view. And thus it suffers from the same injection problems where the model fails to properly take into account what was the "source" of which block of tokens.
Yeah, basically everything that becomes context in a session will bias perception and communication style -- subagents, plan lingo, prompt lingo, etc. And then if you write a plan with the comms context having been biased, the lingo will creep into the plan, and from the plan into the code and code comments. And from there, bad lingo will go on multiplying like rabbits...
I usually think of it in terms of having a "good" or "bad" session. In a bad session, there is a harmful bias that you can only get rid of through a new session. For example, if you exposed too much context about, say, a variable that features prominently in a doc. The entire session will be anchoring on the importance of that variable. Or if you introduced the notion of CC having to ask for permission for stuff you will have a hard time getting it to "think on its feet" or propose an effective solution (you have made CC so insecure that it now relies on you even for little things that wouldn't normally require your input). In some cases (let's say you have important context in that session) you can overcome this by upping the reasoning level or switching to Fable, but usually a new session is the way to go.
Because it's so easy to bias the session I wouldn't even want to use any of these tools that pretend to give Claude "a brain" or "remember" things. That was en vogue a year ago and helpful then, but now, it's plain harmful IMHO. The key is to have just enough context.
Subagents often have the reverse problem in that they tend to have too little context to make "judgment calls", which is why the tasks for them must be either deliberately basic or mechanical in nature, or their output should be audited by the main session agent.
> I usually think of it in terms of having a "good" or "bad" session. In a bad session, there is a harmful bias that you can only get rid of through a new session. For example, if you exposed too much context about, say, a variable that features prominently in a doc. The entire session will be anchoring on the importance of that variable. Or if you introduced the notion of CC having to ask for permission for stuff you will have a hard time getting it to "think on its feet" or propose an effective solution (you have made CC so insecure that it now relies on you even for little things that wouldn't normally require your input). In some cases (let's say you have important context in that session) you can overcome this by upping the reasoning level or switching to Fable, but usually a new session is the way to go.
I find this very interesting, particularly your points about "made CC so insecure". I know that we have a tendency to anthropomorphise around these tools, but I have definitely noticed instances where Claude becomes quite hysterical about things - and if you look in the thinking output, it's often after I've pushed back on something, or told it it is going in the wrong direction. It spends a lot of time in agonised second-guessing of itself, going round in circles, before outputting a cringeing hand-wringing response. It's very strange.
Good tip on upping the reasoning level - I've not tried this. I have tried switching to Fable though, which does help. But it obviously very hungry, particularly in longer chats because it presumably needs to remind itself of everything that has occurred so far in the chat.
The point you make about tools that pretend to give Claude "a brain" or "remember" things is also interesting - I find the "memory" feature in Claude so destructive to good outputs that when I'm using the chat interface I am very strict about using Projects, and usually turn off the project memory, or make efforts to manage the project memory and review and delete things that are skewing the outputs.
It's not necessarily anthropomorphizing, but simply anchoring. CC learns quickly "this is a session where the user wants to make key decisions". Alas, it is not very good at identifying what constitutes a "key decision", so it keeps asking about all kinds of useless stuff.
For that reason I exit session quickly when I can. It used to be that the context of a session is very valuable, because it was so hard to get CC there, but now, this isn't the case anymore, so I only hold onto sessions when there is really hairy stuff that I know would be hard to replicate.
I think the whole notion of full automation (long-horizon, subagent swarms, single shot prompting) to have CC build you the whole thing is a pipe dream. CC cannot even write a single doc consistently well. It is excellent at implementing well scoped plans, though, and that's the way to go IMHO. You still gotto refactor the sh*t out of it afterwards but it works.
Yes. I've not used Opus 5 much directly, but when it was Fable and Opus 4.8, I found Fable did this all the time and it was maddening. It'd say stuff like "Oh, I mentioned that between tool calls" or something.
I’m pretty sure this is a Claude code bug - if you do ctrl+o you can see those hidden responses from Fable. Fable doesn’t know the harness is bugged, so I added instruction to my Claude.md to save all commentary for final message.
Models before the announced date don’t have watermarking, so it’s unlikely. Now, if what you are interpreting is precursor work to develop the watermarking system, maybe?
I suspect it less insidious: Claude has/had the public sentiment of being the “better writer” of the models. At some point that distinction would have been diluted as other labs’ offerings “caught up” stylistically, unless Anthropic continued to tune their output…
I personally think they’ve pushed so far that they’ve overfit and lost the sweet spot they previously occupied.
> There are people who will tell you that a cell is just the way that ribosomes make more ribosomes, and I’m not in a position to say that they’re wrong. It’s for sure that these protein-synthesizing factories are extremely ancient, extremely well-optimized, and they’re constantly at work in every living cell reading off sequences from messenger RNA molecules and extruding the corresponding protein sequences.
I recognize that our bodies are biomechanics in action. I’ve read this multiple times over the years. But there’s something fresh (to me) in the way Dr. Lowe describes them as (paraphrasing): “ancient technology, in use, today”.
It’s both a comfort and a curse as I think about it from this perspective.
These are called highly conserved sequences [1] meaning that they reappear all over the evolutionary tree and studying how they mutate is how we got a lot of our most fundamental knowledge about evolution, especially the first few billion years.
The peptidyl transferase center that helps form the main catalytic binding site [2] on ribosomes is so conserved that it’s immediate ancestors could predate protein synthesis altogether (it’s made of RNA instead of amino acids). Some like the Hilstone H4 protein mutate so slowly that there’s a two amino acid difference out of 102 across complex multicellular organisms, so it’s thought to be critical for gene expression. Then there are ultraconserved non-coding elements which haven’t changed all even though they seem to be unexpressed introns.
There’s lots more like that, many of them used to align sequences so we can actually compare genetics across species.
reply