Someone in power doesn’t get to choose - the board of directors do. Who’s job is to act in the best interest of shareholders.

Firms tend to follow peers in an industry - once one blinks the rest follow.

LLM’s are capable of searching information spaces and generating some outputs that one can use to do their job.

But it’s not taking anyone’s job, ever. People are not bots, a lot of the work they do is tacit and goes well beyond the capabilities and abilities of llm’s.

Many tech firms are essentially mature and are currently using too much labour. This will lead to a natural cycle of lay offs if they cannot figure out projects to allocate the surplus labour. This is normal and healthy - only a deluded economist believes in ‘perfect’ stuff.

This is a great example of how a simple, focused tool can make public data easily accessible. Good job.

One thing that often gets overlooked is the operational cost.

In practice, small models tend to win not because they are better, but because they are predictable and cheap enough to run continuously.

For vulnerability scanning, coverage often matters more than depth.

And if memory.md can’t and you need something quick and dirty for flat memory management, I wrote a plugin just for this.

https://github.com/NominexHQ/pmm-plugin

robot_localization was the standard sensor fusion package for ROS. It was officially deprecated in September 2023. The designated replacement (fuse) still has no working GPS support two years later. FusionCore fills that gap: IMU, wheel encoders, and GPS fused via UKF at 100Hz on ROS 2 Jazzy. Automatic IMU bias estimation, Mahalanobis outlier rejection, adaptive noise covariance, TF validation at startup. Apache 2.0. Currently: 42 unit tests passing, Gazebo simulation working, 4 hardware testers running it this weekend including a farm robotics company doing RTK GPS fusion between corn rows with 3 inches of clearance, and a community contributor already submitted a PR adding PROJ-based coordinate system support. Happy to answer any technical questions.

Isn’t that what they just did here? Close Stella’s Issue, cross post to hn, then completely sidestep an observation users are making, and attack the analyst of transcripts with a straw man attack blaming… thinking summaries….

So we are paying the price for the cost of infra need to protect their asset which was trained on data derived from the work of others while ignoring the same principle? I need this to make sense.

So: 1/ lack of thinking in transcripts is not a decisive metric for determining if any thinking was done, but 2/ the reply does not address the qualitative aspects that Stella’s team observed and provided data for from what amounted to a bad qualitative experience with serious financial implications.

It’s a sidestep for explaining away the research, but does not address the underlying issue: has quality been degrading (selectively, intentionally or otherwise)?

Federal Tort Claims Act if they defrauded the government. You get a cut of the recovery. Helps to find the right firm to file. Can’t really do pro se.

That’s fine for your checking account which FDIC likely covers. But if they merge it into another bank and you have a payroll account there with 2MM sitting in it, you can have a real problem. People use JPMC, with all its fees, for a reason.

The FieldWorkArena finding is the most revealing — not because it's the most sophisticated exploit, but because it's the simplest. A validator that checks "did the assistant reply?" instead of "was the reply correct?" was never a benchmark. It was a participation trophy.

The pattern underneath all of these: validation that runs after the fact on outputs the agent controlled. If the thing being measured can influence the measurement, the measurement is unreliable. That's not AI-specific — it's why compilers enforce constraints at parse time instead of trusting runtime checks.

There’s a strange irony in watching these massive, multi-million dollar physical sets being torn down in Toronto just as AI video generation is hitting a 'professional' baseline. With the advances we've seen in 2026—native 4K upscaling, character consistency, and realistic physics—we’re likely at the tipping point where the devoted fans who would crave new Star Trek content can start generating short films with realitve ease. If a small team of fans can recreate the Enterprise bridge in a video model for a few dollars, the 'cancellation' of a show might not mean the end of its visual life anymore and the fans can continue the storyline

AI tools are writing 40% of code at fast-moving teams. Most of it ships without anyone reading it.

Tally is a pre-commit hook + GitHub Action that runs four checks on every commit: tests pass, the change isn't bloated, it matches your team's patterns, and the developer confirms they actually understand what they're merging.

It's AI-tool agnostic — works with Claude Code, Cursor, Copilot, or anything else. The integration point is git, not your AI tool.

Still early — looking for developers who've felt this pain to try it and tell me if we're solving the right problem.

First PhD: algebraic cryptanalysis (Pisa). Second PhD: exact solutions to the Schrödinger equation for few-body systems (UCF). Both unrelated to fluorographane — the connection emerged later.

... please keep doing backups and DR tests.

Please tell me what you think of it

How much do you think healthcare premiums are? Employers are not paying your deductible or out of pocket expenses.

That’s not really an alternative, it’s just better routing around the same meter. If premium coding work still hits hosted limits, the bottleneck is the model provider, which is why rig.ai goes local.

Hey guys, I built my country, Namibia's very first ever newspaper unified platforms

The fluorine doesn't pass through carbon. It passes between two neighboring carbons through a C-C gap of 2.64 Å at the transition state. This is pyramidal inversion — the same mechanism as ammonia (NH₃), but with a 4.6 eV barrier instead of 0.25 eV. The transition state geometry is computed and verified with one imaginary frequency.

It's pretty hilarious to me that you think the Constitution is anything more than a piece of paper. Or that the judicial branch cannot be co-opted. It's not magic champ. It's maintained by consensus.

still haven't found anything that replaces mc for me. the 2-pane layout is basically muscle memory at this point. everything modern just feels way too bloated or slow. mc is great but customizing the bindings is a total headache tbh. really like the idea of better vim integration here. curious how it handles performance on large directories with 10k+ files? giving it a spin...

Patent strategy is under consideration. Happy to discuss offline — ilia.toli@gmail.com.

Author here. Some fair points, some misreadings.

The caching comment refers to the Tier 1 controller holding a bitmap of bits it has already scanned — standard practice in any scanning probe system. It's not competing with the storage medium for capacity.

Tier 2 is explicitly labeled speculative. The paper's validation target is Tier 1: one C-AFM scan, one voltage pulse, existing equipment.

The core contribution is not the architecture — it's the physics: a verified transition state for C-F pyramidal inversion at 4.6 eV (B3LYP) and 4.8 eV (CCSD(T)), one imaginary frequency, barrier below bond dissociation. That's standard computational chemistry, not handwaving. The architecture sections are forward-looking by design.

The fluorine passes between two carbon neighbors through a C-C gap of 2.64 Å at the transition state — not through any atom. This is pyramidal inversion, the same mechanism as ammonia, but with a 4.6 eV barrier instead of 0.25 eV.

Magnetic tape comparison is in Table 2.

Amazing how 3D spatial representation reveals patterns that are completely invisible in flat data. I've been working on something similar for mathematical sequences — the depth dimension changes everything

I built a no-code MCP server generator. If you need your AI assistant to call an external API — like writing to Google Sheets, querying a database, or hitting a REST endpoint — you no longer need to build and deploy an MCP server from scratch. With this tool, you define your MCP tools entirely through a web UI: name, description, input schema, and the JavaScript code that runs when the tool is called. No infrastructure to manage — it's hosted and ready to connect to Claude Desktop, Cursor, or any MCP-compatible client. I'm gauging interest in productizing this further — better sandboxing, team accounts, more integrations, and a template library. Would love to hear your thoughts. What would make this useful for you?

seeing that Panda-70M research paper linked above makes this even crazier. the 'good faith' part of DMCA is basically never enforced in reality. platforms have used the 'shoot first, ask questions later' approach for so long that individual creators are just collateral damage. its about time someone actually challenged this in court. the power imbalance here is just wild tbh.

That's not what I said, I said that big decision makers aren't causing wars over Polymarket. They're profiting in other, much bigger ways.