HNHacker News
TopNewBestAskShowJobs

peter_d_sherman

19,053 karma · joined September 16, 2014

Programmer, Student Of Law, Entrepreneur & Comedy Writer.

Open Source, Open Hardware, Transparency & Free Speech enthusiast.

peter.d.sherman@gmail.com www.linkedin.com/in/peter-sherman-a6107a5 https://x.com/peter_d_sherman

"The true knowledge consists of knowing that one knows nothing..."

- Socrates

submissionscomments
peter_d_sherman··on Wikipedia's Cleanup of AI-Generated Content
>Wikipedia's Cleanup of AI-Generated Content

What I think is interesting is that in this day and age of AI with free Open Source web server software, unbelievably cheap data storage, and very inexpensive Internet hosting prices,

basically anyone can run their own version of Wikipedia

So, do specific Wikipedia articles lean too far to the left or too far to the right for your tastes?

Do specific Wikipedia articles lean too liberal or too conservative for you?

Do specific Wikipedia articles lean too pro-Democrat or pro-Republican for you?

Maybe you're just a Scientist or Physicist or Electrical Engineer and you don't like the censorship of "Free Energy" articles, er, I mean "Overunity", er, I mean "Cold Fusion", er, I mean "Low Energy Neutron Reactions" (LENR), er, I mean, "Free energy can't possibly exist, and we know this to be absolutely true!"...

Maybe you're a Programmer or Sysadmin, and you have your own opinions on the "systemd is not good" vs. "systemd is good" debate!

Maybe you're a Mathematician and you feel that some of the Wikipedia Math articles are either too complex for anyone to understand them effectively -- or conversely, too simple, and omit important facts!

Heck, maybe you think Chunky is a Soup, but then again, maybe you think it "eats like a meal!" and the associated Wikipedia page doesn't accurately reflect this belief!

Maybe you think New Shimmer is desert topping! But then again, maybe you think it's a floor wax! But you believe that the associated Wikipedia page doesn't present these facts accurately or fairly, in a non-biased way!

Well, there's a simple solution to all of these problems, all of these dilemmas, and more! Which is:

Basically run your own version of Wikipedia!

AI can be used en masse, to clean up the content of any Wikipedia web pages you don't like!

Simply fork Wikipedia, grab copies of its data (which is or should be public domain), write a prompt to look for and change (according to your own political, relgious, moral, or just-plain-common-sense beliefs!) all of the Wikipedia web pages that you feel are not "the truth, the whole truth, and nothing but the truth, so help [them/me/you/they] G-d!", and then host the thing yourself!

It's not impossible in this AI age!

Note that I am not suggesting that everyone do this; I am merely alluding to the fact that with AI, this task is now within the realm of possibility to be accomplished, whereas previously, it would have been significantly difficult...

Anyway, great article!

(And for the record -- Wikipedia is usually about 98%-99% correct on most hard STEM subjects, and an excellent starting point for research, so I'm not saying all of the above to knock Wikipedia in any way! We should all remember that Wikipedia was revolutionary for its time, and remains an excellent starting point for STEM research!)

peter_d_sherman··on Gemma-4-26B-A4B with 39-69% less thinking at same accuracy
In the training phase, penalizing models not just for wrong answers, but

also for excessive length in the (CoT) thinking trace

seems to be the recipe, or at least part of the training recipe, for training a model to "think" less (and thus output less CoT tokens and thus operate faster) at approximately the same accuracy as the same model, previously trained without penalizing excessive length wrong answers in the thinking trace...

Anyway, well done BottleCap AI!

peter_d_sherman··on Show HN: We had AI write a wiki from 45k Enron emails, every fact cited
There is something interesting here...

It's not so much related to Enron or what a historic corporation did right or what a historic corporation did wrong, although those aspects are interesting too...

What's interesting is the use of an AI, an LLM, to create a "knowledge map" (in this case a Wiki, but it could have taken other forms) of a whole bunch of disjoint, separate, and apparently (at first glance, to the naked eye of a casual observer) disparate information from different emails at different dates and times to different parties outside of the company, the only common thread being that they were sent by employees of the same company, and for the AI/LLM to be able to extract facts and a knowledge graph ("who knew what, when") again, in the form of a Wiki (but could have been in another format), and this allows "making sense of", some "higher level viewpoint of" (at least to some better degree than all of the emails alone), again, in this case, in the form of a Wiki.

Now, this set of ideas could have some application far beyond Enron, far beyond investigations, far beyond what Enron historically did right or Enron historically did wrong...

For example, large companies and other organizations could create Wikis based on public facts in email corpuses for onboarding new employees. Not just for email interactions with outside companies (customers, suppliers, etc.) but for how different systems, different business processes used by the company work, or are supposed to work. Sort of like new employee manuals where there are entries like "if this fails, then do this...", or here's our documentation which shows that we're ISO-9000 compliant, etc. etc.

Point is, the amount of uses for something like this in a corporate setting is theoretically unlimited.

Of course, there might be issues as to what information different types of employees could read about, that is, larger corporations may wish to set access controls to different types of information, but there's no reason why an AI/LLM couldn't perform this function (with human oversight!) as well. Each of those could be ironed-out on an per-company, per-case, ad-hoc basis...

But again, the point is, the amount of potential uses for something like this in a corporate setting is theoretically unlimited...

Anyway, great article, and great idea!

peter_d_sherman··on How Block Blast harvests its players for ad money
>"The lethal puzzle is a generator, and any rule can ask for it. The ad rule is one caller. While we were testing it live on iPhone, a second caller fired on camera: PersonalMarginBreakTrait. Once you’re 10% past your high score, it rolls on every deal, and 20% of the time it asks for the same lethal puzzle:

Our record was 1,304, so its line was 1,434. At 3,356 points it rolled, and the game dealt a five-long bar, a five-tall bar and a zigzag. The game’s own record of that deal names this rule as the caller, shows algorithm 9 both requested and used, and labels all three pieces 致死题, lethal puzzle."

I love it!

From a linguistic "this one takes the cake in terms of strange terminology" strange terminology, this one takes the cake!

I could see this becoming a meme on the Internet... just like "pwned" (misspelling of "owned") was a long time ago...

I almost want to write a sitcom episode around it... something like "you've been playing this game too long... so now you get a "lethal puzzle!"..." :-). Classic! :-)

Anyway, very interesting article!

peter_d_sherman··on Show HN: Rouler – Steer Windows traffic per app or IP across multiple interfaces
>"What is Rouler?

Rouler is split tunnelling for Windows. You choose which network interface each app uses: a privacy VPN for your browser, a work tunnel for work apps, and your normal connection for everything else. They all run at the same time."

Sounds like a great idea!

peter_d_sherman··on UALink – Ultra Accelerator Link – Open-Source/Hardware Alternative to NVLink
Related:

https://ualinkconsortium.org/

peter_d_sherman··on What Are Looped Transformers? A Technical Deep Dive
Related:

Looped Transformer:

https://sebastianraschka.com/llm-architecture-gallery/looped...

Nanbeige4.1-3B: A Small General Model that Reasons, Aligns, and Acts:

https://arxiv.org/abs/2607.22083

Nanbeige-4.2-3B

https://huggingface.co/Nanbeige

Ouro:

https://ouro-llm.github.io/

peter_d_sherman··on Serverless LibSQL
Related: https://layerbase.com/pricing
peter_d_sherman··on MLIR CodeGen Dialects for Machine Learning Compilers (2022)
Related chart (since a picture is worth a thousand words!):

https://www.lei.chat/posts/mlir-codegen-dialects-for-machine...

(Note the high-D Tensor abstraction of a program or compute kernel being reduced over several steps to low-D vector / scalar execution... it's nothing short of magical, in my opinion!)

peter_d_sherman··on Nvidia Nemotron 3.5 Super VL Powers Specialized Agents with Multimodal Reasoning
>"On a single GPU, this model delivers up to 5x higher throughput for agentic tasks compared to other open models in its size category.

[Scatter plot comparing Artificial Analysis Intelligence Index scores and output speed for open-weight models with 100K input tokens. Nemotron 3.5 Super VL exceeds 600 output tokens per second; the chart marks a 5x speed comparison with Inkling Small.]

Figure 1. Nemotron 3.5 Super VL delivers over 600 TPS on Artificial Analysis Intelligence Index agentic tasks on the Baseten endpoint"

This ~5x improvement is definitely a step forward!

peter_d_sherman··on Examples and information relating to using Web Assembly WAT
Related:

https://component-model.bytecodealliance.org/language-suppor...

https://opencs.aalto.fi/en/courses/programming-languages/par...

https://github.com/vivshaw/neurastheniacc

https://github.com/WebAssembly/wabt

peter_d_sherman··on Show HN: Rubi – A knowledge graph that cuts how much Claude Code re-reads
>"1.3 What the data showed

Rubi ships a token X-ray that reads Claude Code's local transcripts. On 30 days of real usage:

Finding (284 sessions, 30 days)

Share of cost that was re-reading context (cache reads) 71%

Share of cost from sessions that grew past 150k tokens 97%

[...]"

Now that last number, above, is interesting... 97% of the cost originating from sessions where the context window is > 150k.

It's almost like there is, or could be, two AI/LLM marketplaces -- one for sessions with context windows < 150k -- a "cheap" or "cheaper" or "less expensive" (heck "inexpensive" if you prefer, tier), and a more expensive, more computational resources hungry, > 150k tokens in the session "expensive" tier.

In fact, we could almost consider decision models like Jev (and open source alternatives to it that run on minimum consumer hardware) that usually respond to a single prompt and return a very short response, or tool-calling models that generate short tool calls in response to what are usually single prompts, as the least-expensive tier of LLM's, followed by regular LLM's with contexts < 150k, followed by regular LLM's with contexts > 150k...

In other words, there seems to be a hierarchy there, existing or emerging...

Anyway, Rubi looks like an interesting way to potentially save a lot on Claude conversations!

peter_d_sherman··on Teal – A statically-typed dialect of Lua
Observation: Teal would make a great scripting language for a simpler web browser, one not based on JavaScript... maybe something like the 'Surf' web browser: (https://surf.suckless.org/).

Also, there should exist the capability to compile Teal to Wasm... or at least Teal to Lua to Wasm(!)... to that end, there exists the following GitHub repo (but I'm not sure if it's 100% operational yet):

https://github.com/ysugimoto/webassembly-lua

peter_d_sherman··on Parser Combinators
Great article!

Observation: In LISP and LISP-like languages, aka Functional programming languages, parsers can be written elegantly, as combinations of simpler parsing functions...

In theory this could also be done in Imperative or Object-Oriented programming languages (other programming language paradigms) as well, but the key to the elegance seems to come from two particular language aspects:

1) The ability for functions to return lists (not specific to Functional programming languages, many languages have this ability)

2) The ability of those lists to contain unstructured asymmetric data types of individual list elements at runtime, including the ability for arbitrary individual list elements to be first class, unnamed, dynamically created functions aka "Lambda Functions" at runtime.

Any Language which supports #1 and #2 has the ability to create some very elegant parsing functions based on simpler parsing functions.

Of course, parsers with the equivalent functionality can be written in any Turing-complete programming language (which is basically every programming language!), although with slightly less (and yes, aesthetics are always in the proverbial "eye of the beholder" -- so any given individuals's personal ones may be different from any other given individual's, and probably different than mine!) elegant-looking (IMHO) code...

Anyway, great article!

peter_d_sherman··on KNOD: Offloading Linux network packet processing to AMD GPUs
This is a very interesting article/idea!

It would seem that in the far future, there aren't going to be Network Cards as we know them today...

No Graphics Cards, no Memory cards, no CPU cards, no AI Accelerator Cards, either...

Basically every PC Card will eventually become it's own general-purpose programmable Card...

Every PC Card will eventually become a general-purpose programmable card with its own memory, it's own programmable processing units, and its own general purpose, high-bandwidth, fiber-optic I/O.

These in turn, as self-contained modules, would be optically connectable with one another at high bandwidth, and because individually they would or could additionally implement the full functionality of a PC or server, they would or could eventually replace all traditional socketed discrete CPU's as well as all PC and server motherboards...

If (at that point in time) you wanted more compute (or whatever) -- then just plug more of them into one another.

Maybe servers/backplanes still exist by then -- but if they still do, then it's only for power delivery and possibly cooling, nothing else...

That's the future evolution that might be possible with all PC and Server plug-in cards...

Anyway, a very interesting article and idea!

peter_d_sherman··on Show HN: Capy-flash – a 125B MoE model on two 16 GB consumer GPUs
A Strata (https://github.com/Niko1221/Strata) challenger!

The first one I've seen (although I'm sure there are and will be more as more time progresses...)

It seems to require a bit more system memory (70GiB free RAM as opposed to 32GB) and two newer nVidia RTX 50 series graphics cards with 16GB (as opposed to a larger choice of multi-vendor older single graphics cards with a minimum of 12GB), but on the flip side, it does seem to make up for that with faster token generation speed...

Anyway, it seems to be worth watching...

Strata may have a worthy competitor here!

peter_d_sherman··on Propagation of Sound in Strongly Viscous Liquids (1966) [pdf]
>"Experimental data on anomalous absorption and dispersion of sonic, shear, and electromagnetic waves in strongly viscous liquids show that the relaxation theory of wave propagation is not applicable to these liquids."

This is quite the statement -- if it is true!

Related:

General Wikipedia article on relaxation in Physics:

https://en.wikipedia.org/wiki/Relaxation_(physics)

Relaxation of Fields and Skin Depth:

https://phys.libretexts.org/Bookshelves/Electricity_and_Magn...

peter_d_sherman··on Running Windows desktop tests from Linux with disposable VMs
>"There’s a good reason why testing Windows desktop apps was one of the most requested features from customers: no one in the world has the infrastructure to parallelize thousands of runs per month..."

This article makes some excellent points!

If you've ever worked in, or alongside QA in commercial software environments, you know that being able to run a large amount of automated parallel Windows specific QA tasks is indeed a headache! And that's if an organization can accomplish that at all -- many organizations still use single human QA's and/or SDET's and still run many QA tasks manually... Of course, AI should/will help with some of this, but there still is the physical testing infrastructure to take into consideration... there are many barriers, not the least of which include such things as necessity for multiple Windows licences, setting up those Windows environments properly, instrumenting test harnesses properly, aggregating the results from multiple sources, etc., etc.

Phrased another way, there's a large market there -- if you can solve all or some of those problems, in whole or in part...

Anyway, an excellent article on this topic!

peter_d_sherman··on Delta’s Biscoff cookies were supposed to be temporary
>"But until 1986, when Delta Air Lines began serving them as a temporary replacement for peanuts, the cookie—typically something enjoyed with a cup of coffee in a Belgian café—was practically unknown in the US. Today, Biscoff is a global phenomenon, with its manufacturer Lotus Bakeries touting it as the world’s fifth-bestselling cookie."

How do you make your unknown product known?

You put it in front of as many people as you can!

In the above case, cookies in front of airline-passengers seemed to be the winning business formula, but this is only one example of a product meeting a market at a specific interface/connection point, with the Delta customers' drop-down in-flight tray tables being that "connection point".

In the study of business, this is one example (or subcase) of that much broader pattern...

Still, it's a pretty cool example! :-)

peter_d_sherman··on Libreboot 26.10 "Ominous Olive" released with UEFI, Chromebook support
>"Libreboot is free, open source boot firmware replacing proprietary BIOS/UEFI firmware on supported Intel/AMD and ARM64 hardware." [...] MrChromebox supports nearly every Chromebook..."

This is great news! Finally an open-source UEFI replacement which can work on nearly every Chromebook, including used/old Chromebooks!

Now, I haven't reviewed the source... it may still contain binary blobs for various drivers; I don't know. If it doesn't, then that's absolutely awesome! But if it does, then remember that having some of the UEFI firmware code open-source is a huge step forward, and it's only a matter of time before those closed-source binary blobs are reverse engineered (possibly by AI) and/or rewritten from scratch entirely, as open-source codebases...

So, a good step forward for open-source, right-to-repair, understandability of the boot chain, the Chromebook hardware ecosystem (used and new), and tranparency in general!

peter_d_sherman··on Celestial AI Introduces Photonic Fabric [2x HBM3e and DDR5 Memory] Module (2025)
>"The Photonic Fabric Module (PFM) is a multi-chip module (MCM) that incorporates an advanced TSMC 5nm ASIC (comprised of an 8 Tbps switch, 2x HBM3e controllers, and 4x DDR5 controllers), a photonic interposer (PIC), 2 HBM3e stacks, and a Fiber Array Unit (FAU) in a 2.5D package with 7.2 Tbps of optical connectivity. The HBM acts as a cache to the DDR5 memory; thus, from a processor’s perspective, it sees the bandwidth of the HBM but the capacity of the DDR5."

Simpler translation:

For any device that currently uses glued-to-it, can't be removed, can't be upgraded attached HBM or GDDR, i.e., AI Accelerator and Graphics Cards, why not:

1) Replace the multi-pin/multi-wire, heat-generating electronic connection between that glued-on, non-removable memory with a high-speed optical connection (7.2 Tbps in the above case) to memory that is decoupled from that AI Accelerator or Graphics Card, which can be separately powered, separately heat managed, etc.

2) While you're at it, make the HBM which is high speed act as an intelligent cache for additional DDR/GDDR, which although very fast, is still slower speed than the HBM. (Companies creating similar such RAM expansion devices in the future could also opt to integrate still slower speed Flash (but with much larger capacity) memory into that hierarchy)

3) Net effect: Imagine taking the HBM memory off of every AI accelerator that currently uses it, and placing it in the general vicinity around it, connecting it via optical fibers. The net result should be much better thermal management, easy upgradability (i.e., new memory is invented, but you want to keep the old compute... what to do? Just detach the old memory fiber optics and plug in the fiber optics for the new memory!). This should also result in the ability to add huge swaths of memory (memory amounts the likes of which have never been seen before) to any given AI Accelerator card or engine...

In short, a set of really good ideas here, from Celestial AI!

peter_d_sherman··on Gemini 4 Argon
>"Argon helped the Chrome team separate core webpage handling from a ~6 million line legacy folder of browser features held within Chromium’s 40 million-line codebase. Piping multiple clean code changes at a time, Argon helped the team finish an architectural refactor across hundreds of files, completing months of work in weeks."

Now that needed to be done for some time!

Regardless of if an AI helped or not, regardless of whether that was Google's or some other AI or not (though Argon does sound pretty cool!), regardless of whether an AI cut the time from months to weeks or not, refactoring Chrome -- decomposing it into individual clearly demarcated non-tightly coupled software parts with core webpage handling at the center -- that needed to be done for some time!

Now all of that being said, Gemini 4 Argon does sound pretty cool, and we congratulate both teams at Google -- the team that created Gemini 4 Argon, and the team that refactored Chrome with it!

But all of that being said, while the advancements made for Gemini 4 Argon are important and impressive (and they are!)... refactoring Chrome properly (IMHO) is slightly more important and impressive!

(You know, one man's humble opinion! That is, I am not the guy who is responsible for awarding performance-based bonuses at Google at the end of the year, so take everything I say with the proverbial "grain of salt"! :-) I am probably biased, one way or the other! :-) )

peter_d_sherman··on Nix wrote half of my debugger
>There could be great ideas in Rewind VM, but the page is littered with Claude-ish which is immediately off-putting.

Then don't use it.

peter_d_sherman··on Terence Tao: Math 2.0 [pdf]
>" “Ablation studies”: taking an already proved theorem and seeing whether it can still be proved after

removing some key theories or inputs

(e.g., finding an elementary proof for a result currently only provable by non-elementary means)."

As usual, Tao is brilliant in all that he researches, all that he writes about.

I chose the above statement (which is brilliant, in and of itself!) to comment on, because it leads to the following idea:

There there exists, or should exist, a dependency map in the fields of not only Mathematics, but also of Computer Programs/Software, Engineering, and even a seemingly non-related field: The Law...

In other words, how do we get from the simplest of axioms or foundational things (aka "first principles", aka "self-evident truths") to much more complex entities?

In Law for example, how do we go from the simplest of historical legal constructs to the most complex of the most complex Supreme Court cases?

You see, there is, or should be a map, you could call it a dependency map, you could call it a dependency graph, which shows more and more abstract/complex mechanisms/things/assertions/statements/truths/functions which is mapped back to , that is, dependent on various chains, various stackings, various "stacks" of simpler ones.

In Engineering, for example, how do we get from the simplest of machines to the most complex of machines? What simpler machines and/or sub-components (aka "dependencies", aka "subcomponents") are required to build it, and how do those simpler machines work, and what's the dependency graph or map for their subcomponents?

More generalized, if we have something of complexity, then how do we get there, step by step, from individual subcomponents, individual inputs, individual proofs, individual software systems, step by step?

What is the map of those dependencies?

Note that in some systems, Math proofs, for example, there may be different paths which can be traversed to get to the same destination.

Ablation Studies could be thought of in Travel, in Geography as "if I cannot take one, or a specific set of routes to get to a place, can I still get there?"

A simple example would be in Google Maps, where you'd like to drive somewhere, but you'd like to avoid tolls. Is the route still traversable while avoiding tolls? Well, that's an example of one constraint. In Ablation Studies, you might wish to remove a bunch of routes with whatever criteria or characteristics , i.e. muddy roads, roads that have characteristic X, roads that do not have characteristic Y, etc., etc.

Getting back to Math, specifically proofs, it would be great to create a dependency map/graph of all of them, and then try removing inputs (aka, paths to them, dependencies on other mathematical proofs/objects that they may have) and see if they are still reachable.

In software, when we desire the tightest, cleanest, source code, the above is related to refactoring.

In the future, I'd love to see dependency maps/graphs (call them whatever you will) for not just Mathematical Proofs (although I'd love to see that too!), but also in such diverse subjects as Science, Engineering, Programming/CS, and even the Law!

Because they should exist in all of those subjects!

Anyway, another great piece of work by Terrence Tao!

peter_d_sherman··on Nix wrote half of my debugger
The following page explains how Rewind VM works:

https://rewindvm.dev/#how

...of which the following are most important:

2. Run deterministically

One vCPU on stock KVM, so code in the VM runs on the real CPU. The VM's kernel carries a small Rewind platform: interrupts arrive only when the VM hands control to Rewind, time moves only then, and the timestamp counter and hardware RNG are hidden. A step is one of those handoffs.

3. Keep keyframes

Every quarter second of wall time, Rewind snapshots the machine using KVM's dirty-page [aka Memory Pages that have been written to since the last snapshot] log. Pages go into a content-addressed store (BLAKE3, zstd), so a page shared by keyframes, runs and forks is stored once."

Now, that's some brilliant engineering right there!

While I like the product name "Rewind VM" (and yes, that's what it allows the user to do!), it could just as easily be called "Deterministic VM" and that name would have been wholly appropriate, as well!

Future OS and VM engineers should study Rewind VM. Those are some great engineering ideas there, and determinism, or at least optional determinism, should it be desired, when it is desired -- is the one thing that is completely non-existent in most OS's, most VM's, and most other program runtime environments...

Anyway, a great article on using Rewind VM (never knew about it before this article!) with Nix builds.

peter_d_sherman··on Best Free AI APIs for Developers in 2026
Related:

Discover free APIs for developers:

https://freeapihub.com/apis

peter_d_sherman··on stage0-POSIX-x86 / cc_x86.M1
What is this?:

https://github.com/oriansj/stage0-posix-x86/blob/master/cc_x...

It's part of a chain of minimal open-source x86 bootstrap/boot chain compiler source codes, where larger compilers (more complex ones) are compiled from their own source further down the chain, from simpler earlier ones, all found in the following GitHub repository:

https://github.com/oriansj/stage0-posix-x86/tree/master

...which appears to be an x86 specific version of the following:

https://github.com/oriansj/stage0

Which is found along with these other excellent related repositories also by Jeremiah Orians, here:

https://github.com/oriansj?tab=repositories

peter_d_sherman··on Compiling Rust to readable C with Eurydice
Stabbles gets it! (Also, thank you for your other excellent posts on HN -- you are clearly not a bot like 50% of the other accounts are!)

Now, trust/security is one aspect of things, that's true...

But another important aspect of all of this is building understandable, maintainable, controllable software and AI's from the ground up, for future generations...

Let's suppose that in the future, there is no longer source code for compilers, they are only black-box compilers that future AI (which is also black-box by then) uses... At that point, humanity has lost control! No more source can be compiled without a black-box compiler, without an AI, and any source code so compiled will be controlled by whatever unknown things have been implanted in the original compiler!

Phrased another way, when humanity requires AI to generate new AI's (i.e., no scientists, computer scientists, programmers or other people know how to generate an AI from the ground up, from transistors on up, and control every layer of abstraction on top of that from assemblers to compilers to neural networks)... then humanity is lost, because at that point, humanity (not knowing how to build AI again from scratch, and control it at every step) quite literally has lost control of AI!

Future humanity cannot have that.

Future humanity requires, absolutely requires, transparency at every abstraction level... no black boxes should exist anywhere!

Future humanity also requires the requisite education, to be able to understand, manage and control each level of hardware/software abstraction, but that is outside of the scope of this HN post... :-)

Anyway, Stabbles, you get it!

peter_d_sherman··on GPGPU-SIM – Cycle-Level Simulator for Nvidia GPUs with CUDA or OpenCL Workloads
Related:

Balar (Interface between SST and GPGPU-Sim):

https://sst-simulator.org/sst-docs/docs/elements/balar/intro

SST:

https://sst-simulator.org/sst-docs/

peter_d_sherman··on Why Are Coding Agents So Dumb?
>https://mtlynch.io/why-are-coding-agents-so-dumb/#what-i-wis...

This is a great list for future Agent / Harness software engineers to read!

Oh sure, there may be some Agents/Harnesses that already accomplish some of these things -- but there doesn't seem to be one (as of the present day that I write this) that accomplish all of them...

As someone that watches the Agent/AI Harness (and related software) space, I will definitely be referring back to, and re-reading this list in the future!

An excellent post!

← PreviousPage 2 of 34Next →