I'm being hyperbolic of course, but I'm a little dismissive of the progress that happened since the days of BBS's and car based cell phones - we just got more connectivity, more capacity, more content, bigger/faster. Likewise, my attitude toward machine learning before 2023 is a smug 'heh, these computer scientists are doing undisciplined statistics at scale, how nice for them.' Then all of a sudden the machines woke up and started arguing with me, coherently, even about niche topics I have a PhD in. I can appreciate in retrospect how much of the machine learning progress ultimately went into that, but, like fusion, the magic payoff was supposed to be decades away and always remain decades away. This wasn't supposed to happen in my lifetime. 2025 progress isn't the 2023 shock, but this was the year LLM's-as-programmers (and LLM's-as-mathematicians, and...) went from 'isn't that cute, the machine is trying' to 'an expert with enough time would make better choices than the machine did,' and that makes for a different world. More so than, going from a Commodore Vic 20 with 4k of RAM and a modem to the latest Macbook.
Might be a good place for yunohost/coolify style services, especially if you have multiple separate entities - though probably tricky to do inbound mail because of IP allocation?
You might start questioning meaning of life with a billion year time budget. A million years seems reasonable to cover the range of things you could anticipate wanting to learn or experience. A few thousand years, no, that's not enough, you have to start cutting corners, you can barely even visit nearby worlds and only cover a few intellectual disciplines.
Garfield was in many ways the most personally appealing and brilliant of the American presidents, rising from poverty and obscurity by being absurdly talented across many fields and eloquent.
He was assassinated early and barely got to serve. The story of his life, the shooting, and the subsequent medical drama (featuring even a cameo by Alexander Graham Bell improvising a diagnostic device) are so epic you have to wonder if time travelers are messing with us.
His legacy was the nonpartisan professional civil service, a key part of his agenda that his successor felt obligated to carry out, an accomplishment that recently came under particularly heavy attack.
Netflix just came out with the miniseries about him, 'Death by Lightning,' based on the book 'Destiny of the Republic.' His earlier life is featured prominently in '1861: The Civil War Awakening' by Adam Goodheart. There are a few great C-SPAN/Book TV videos by some of the authors that tell the story concisely and convey why some of us are so fascinated by that history.
Now if only the Julia community didn't keep insisting on ligatures and impossible-to-look-up Unicode symbols - the hollow semicolon for compose right to left, seriously?
(off topic) The code chunks in the article use a ligature font, so ">=" is rendered in a way that makes you stop to think how to type it - which is especially confusing since the context is not exactly math. Down with ligatures and extra cognitive load!
For math/data-sci/ML, multiple (GIL-bound) interpreters per process, with (unsafely) shared data structures, would get us much of the way there - basically multiprocessing without the marshalling and process overhead, at the price of pinky-swearing we won't mutate the shared data. That would enable, for example, calling Python-world utilities in-process from a properly multi-threaded language, which can be used to bootstrap e.g. the Julia or Golang (or C#, or Rust...) ecosystems with all the math that's now locked into Python world. If we want NumPy/Scikit etc. to be accessible via thin wrappers, it's tolerable if the Python layer is slow, but importing the GIL into the host language is too high a price.
If your goal is to prove what an awesome writer you are, sure, avoid AI.
If your goal is to just get something done and off your plate, have the AI do it.
If your goal is to create something great, give your vision the best possible expression - use the AI judiciously to explore your ideas, to suggest possibilities, to teach you as it learns from you.
SEEKING FREELANCER | Seattle, Remote fine, any timezone
Senior Role:
C# and Python: If you have XAF experience (and preferably already have a DevExpress license?) AND some LLM API (Semantic Kernel or direct) experience, talk to me about some mainstream LOB work. WinForms and Blazor, document processing/OCR workflow, light deployment/devops tasks, integrating Python and C# based workflows.
---
Junior Role:
Front-end JS (including ThreeJS/Babylon) and light Python backend work, LLM API's.
Set up cloud services, create/maintain a web/social media presence, some marketing/'growth hacking' tasks, writing docs/consumer text.
Probably a good fit for someone who's working on their own micro-SaaS.
---
Please include an hourly/weekly rate, and a one-sentence mention of what your current/preferred AI coding tool workflow is. sep2025-contractor@waldrews.anonaddy.com
Doesn't look like they blew up the API use cases, just the consumer UI access. I wouldn't be surprised if they allow it again, hidden behind a setting (along with allowing the different routed GPT5 levels to be in the selector).
Does caching make as much sense as a cost saving measure on Cerebras hardware as it does on mainstream GPU's? Caching should be preferred if SSD->VRAM is dramatically cheaper than recalculation. If Cerebras is optimized for massively parallel compute with fixed weights, and not a lot of memory bandwidth into or out of the big wafer, it might actually make sense to price per token without a caching discount. Could someone from the company (or otherwise familiar with it) comment on the tradeoff?
Your point holds, but the example of Hellenistic Greek seems ill-chosen to make that case - they had horse races and calendars, and mythological mage equivalents that would be reasonable to name a horse after, so the only thing left to map is 'American' as a geographic proper name and 'an important race in America' - which is about as translatable as it gets. Maybe if we pick one of the many cultures that never had horses, their translation would have to throw in so much context that the corresponding text is structurally different?
The append-only, text csv format you can concatenate to from a script, edit or query in a spreadsheet, and that's still fast because of the in-memory pointer cache, seems like a big win (assuming you're in the target scaling category).
The model that's doing AI Summary for search results - that presumably needs to be fast and cheap because of the scale - is still sufficiently bad as to give people a bad taste. Presumably they're frantically working to scale their better models for this use case. If you could get Gemini Pro on every search result the experience would be effectively perfect (in the sense of better error rates than what a non-specialist educated human reading the top results and summarizing them would achieve). That's years away from a scaling/cost/speed perspective.
And while we're at it, can we do teleportation too? Think of all the parking hassles, the car accidents, the carbon emissions we could save! And yet, just try to convince those funding agencies...
One is a technical advance as important as anything in human history, realizing a dream most informed thinkers thought would remain science fiction long past our lifetimes, upending our understanding of intelligence, computation, language, knowledge, evolution, prediction, psychology... before we even mention practical applications.
The problem is that you can't host a free LLM based service the way you can host a website, without being exposed to cost spikes the moment it becomes popular (or misused). Lots of smaller apps need a better cost pass-through mechanism; this is even more of a problem for hobbyists/non-profit projects than for commercial ones. We can't keep going with a free trial (costs eaten by developer) + subscription for every little thing.
Looks great. We should've stuck with BBS's and Gopher all along. Minor issue, scroll bars are showing and getting slight scrolling when using arrow keys.
I was going to say "so you believe the LLM's don't have the capacity to understand" but then I realized that the precise language would be something like "the presence of photons in this human's retinas in patterns encoding statements about LLM's having understanding correlates to the activation of neuron signaling chains corresponding to, and emission of, muscle activations engaging keyboard switches, which produce patterns of 'no they don't' with high frequency."
The critiques of mental state applied to the LLM's are increasingly applicable to us biologicals, and that's the philosophical abyss we're staring down.
There are type system gotchas. My favorites have to do with integers in place of JS's double-only numerics, the treatment of undefined and nulls, especially in arrays.
The LLM's are just fine at AST translation, though they might inject their quirky preferences if you don't watch them carefully. Suggestion: use the LLM's, but interactively, to translate one script at a time, starting with the simpler ones. Tell the LLM to explicitly call out potential issues. Manually review each, and incorporate the issues and preferences you learn into the prompt. If all goes well, the process will quickly converge on a cut and paste job, but don't be tempted to fully automate it if it's a hundred scripts -- different matter if it's thousands.
Ah, so if doing this once can net tens of millions, all we have to do is repeat the process a few million times, and the national debt is covered. That's where automation comes in.
The smart LLM's are great at this (Gemini Flash seems accurate and cheap), but they can't be trusted not to engage in unexpected censorship, typically skipping parts they find objectionable without reliably telling you that that's what they did. That's annoying enough if you're dealing with, e.g., names that happen to spell something awkward, but it's a big problem if you're scanning medical notes or something else where the awkward text is legitimately needed.
Anyone have success with prompting them to "just give me the text verbatim?"