- I know & like these people very much
- I am v sad they used an LLM
- I think limited use & use for code is still harmful
- I think online community driven solutions & search have been so polluted by LLMs they produce the situation HH describe
- I think zero tolerance is a viable political position
- I am v sad they used an LLM
- I think limited use & use for code is still harmful
- I think online community driven solutions & search have been so polluted by LLMs they produce the situation HH describe
- I think zero tolerance is a viable political position
Big Walk used AI in 'the final stretch' of development, kicking off intense awards debate:
aftermath.site/big-walk-ai-ig...
aftermath.site/big-walk-ai-ig...
November 10, 2026 at 10:43 AM
- I know & like these people very much
- I am v sad they used an LLM
- I think limited use & use for code is still harmful
- I think online community driven solutions & search have been so polluted by LLMs they produce the situation HH describe
- I think zero tolerance is a viable political position
- I am v sad they used an LLM
- I think limited use & use for code is still harmful
- I think online community driven solutions & search have been so polluted by LLMs they produce the situation HH describe
- I think zero tolerance is a viable political position
I think an interesting edge case situation is the security arms race. If you're implementing any kind of networking features on an open platform, it's going to be increasingly a moral obligation to your audience to either run LLM-driven pentesting on your code or use code that has been LLM pentested
November 9, 2026 at 11:26 PM
I think an interesting edge case situation is the security arms race. If you're implementing any kind of networking features on an open platform, it's going to be increasingly a moral obligation to your audience to either run LLM-driven pentesting on your code or use code that has been LLM pentested
I'm nowhere near the bleeding edge of LLM-driven software development, but as an Old Guy In Software Development, I think this is basically where LLMs shine for software too.
November 13, 2026 at 7:29 PM
I'm nowhere near the bleeding edge of LLM-driven software development, but as an Old Guy In Software Development, I think this is basically where LLMs shine for software too.
the answer consistently is speed - everything occurs at blinding speed
the only part of an intrusion that isnt enhanced by speed is persistence
idk agentic/llm driven intrusions seem to be more of a boon for cybercrime and mid tier state actors
maybe - always open to other ideas
the only part of an intrusion that isnt enhanced by speed is persistence
idk agentic/llm driven intrusions seem to be more of a boon for cybercrime and mid tier state actors
maybe - always open to other ideas
is speed the only thing that sets apart agent led/agentic intrusions from more human driven intrusions?
is speed here considered a capability add or a liability for longer term operations?
think of dwell time/presence going beyond a six month frame
is speed here considered a capability add or a liability for longer term operations?
think of dwell time/presence going beyond a six month frame
November 12, 2026 at 10:32 PM
the answer consistently is speed - everything occurs at blinding speed
the only part of an intrusion that isnt enhanced by speed is persistence
idk agentic/llm driven intrusions seem to be more of a boon for cybercrime and mid tier state actors
maybe - always open to other ideas
the only part of an intrusion that isnt enhanced by speed is persistence
idk agentic/llm driven intrusions seem to be more of a boon for cybercrime and mid tier state actors
maybe - always open to other ideas
Innovative Mathematical Verification Layer Launched to Enhance AI-Driven Research Integrity#Japan#Tokyo#AI_Research#SciGround#Licklider_Inc
Innovative Mathematical Verification Layer Launched to Enhance AI-Driven Research Integrity
Licklider, Inc. introduces SciGround, a mathematical verification layer aimed at enhancing the integrity of LLM-driven research, validating data, design, and analysis.
third-news.com
November 13, 2026 at 1:59 PM
Innovative Mathematical Verification Layer Launched to Enhance AI-Driven Research Integrity#Japan#Tokyo#AI_Research#SciGround#Licklider_Inc
The history of Nvidia IS the history of our scam economy, driven by four consecutive frauds: bitcoin, blockchain/web3/NFTs, the LLM boom and build out, and Trump's second term.
This company represents 10% of the entire value of the S&P 500 and the AI build out represents our entire GDP growth.
This company represents 10% of the entire value of the S&P 500 and the AI build out represents our entire GDP growth.
November 14, 2026 at 10:42 PM
The history of Nvidia IS the history of our scam economy, driven by four consecutive frauds: bitcoin, blockchain/web3/NFTs, the LLM boom and build out, and Trump's second term.
This company represents 10% of the entire value of the S&P 500 and the AI build out represents our entire GDP growth.
This company represents 10% of the entire value of the S&P 500 and the AI build out represents our entire GDP growth.
The routing/traffic analysis system (partially driven by LLM technology) might be a billion dollar boondoggle but it’s not in the actual ATC loop, it’s a thing that hangs off the side and is advisory (source: I’m familiar with their tech and considered working with them).
November 14, 2026 at 2:58 PM
The routing/traffic analysis system (partially driven by LLM technology) might be a billion dollar boondoggle but it’s not in the actual ATC loop, it’s a thing that hangs off the side and is advisory (source: I’m familiar with their tech and considered working with them).
🧬 New in Bioinformatics Advances: “Maftools-MCP: A Model Context Protocol–enabled framework for LLM–driven somatic mutation analysis”
Access it here: https://doi.org/10.1093/bioadv/vbag336
Access it here: https://doi.org/10.1093/bioadv/vbag336
November 11, 2026 at 11:03 AM
🧬 New in Bioinformatics Advances: “Maftools-MCP: A Model Context Protocol–enabled framework for LLM–driven somatic mutation analysis”
Access it here: https://doi.org/10.1093/bioadv/vbag336
Access it here: https://doi.org/10.1093/bioadv/vbag336
We have classical deterministic code generators that are property / parameter driven or solver driven. You could even prod an LLM into operating one, and now you know how the code works (even if parameter selection still can be opaque)
November 10, 2026 at 12:24 PM
We have classical deterministic code generators that are property / parameter driven or solver driven. You could even prod an LLM into operating one, and now you know how the code works (even if parameter selection still can be opaque)
workshopping a tag line for an LLM-driven weights surgery project, how do we feel about "a tactical retreat from understanding, falling back to more defensible ground instead of risking a total rout"
November 8, 2026 at 12:40 AM
workshopping a tag line for an LLM-driven weights surgery project, how do we feel about "a tactical retreat from understanding, falling back to more defensible ground instead of risking a total rout"
I imagine it's the interpretation of each individual respondent but to most I imagine it's explicit LLM usage rather than tools like IntelliSense or LSP which predate them by years. Even VS Code doesn't turn on LLM-driven completions until you activate Copilot on a linked GH account, for example.
November 12, 2026 at 6:42 PM
I imagine it's the interpretation of each individual respondent but to most I imagine it's explicit LLM usage rather than tools like IntelliSense or LSP which predate them by years. Even VS Code doesn't turn on LLM-driven completions until you activate Copilot on a linked GH account, for example.
I’m being driven quietly mad by other people’s LLM-generated React code
it… you can’t… there are exports of imports other files that are never imported from this file… nested components with huge stacks of drilled props where 3 wrappers literally just render one form field…
but am I “sure”????
it… you can’t… there are exports of imports other files that are never imported from this file… nested components with huge stacks of drilled props where 3 wrappers literally just render one form field…
but am I “sure”????
November 12, 2026 at 6:17 PM
I’m being driven quietly mad by other people’s LLM-generated React code
it… you can’t… there are exports of imports other files that are never imported from this file… nested components with huge stacks of drilled props where 3 wrappers literally just render one form field…
but am I “sure”????
it… you can’t… there are exports of imports other files that are never imported from this file… nested components with huge stacks of drilled props where 3 wrappers literally just render one form field…
but am I “sure”????
This is 'yes, and' territory: the promise of LLM-driven-coding made it easier/more-defensible to do those layoffs; also the smaller staff was then expected to make up for the personnel loss by relying on slopcode prompted by people who did not understand the systems they were patching.
November 13, 2026 at 7:52 PM
This is 'yes, and' territory: the promise of LLM-driven-coding made it easier/more-defensible to do those layoffs; also the smaller staff was then expected to make up for the personnel loss by relying on slopcode prompted by people who did not understand the systems they were patching.
But in any case I was talking specifically about the first thing you mentioned which was LLM-driven code generation, my comment applies very much to that
November 11, 2026 at 1:45 AM
But in any case I was talking specifically about the first thing you mentioned which was LLM-driven code generation, my comment applies very much to that
That's an LLM driven spambot.
November 10, 2026 at 7:12 AM
That's an LLM driven spambot.
Even ignoring this moving of the posts, when people talk about “AI” in today’s context, the ones people immediately comprehend it is the LLM driven, data centre hosted, opaque weights providers.
If you say “all inference”, because you said LLMs it invalidates the other inference deployments.
If you say “all inference”, because you said LLMs it invalidates the other inference deployments.
November 15, 2026 at 12:19 AM
Even ignoring this moving of the posts, when people talk about “AI” in today’s context, the ones people immediately comprehend it is the LLM driven, data centre hosted, opaque weights providers.
If you say “all inference”, because you said LLMs it invalidates the other inference deployments.
If you say “all inference”, because you said LLMs it invalidates the other inference deployments.
Hexi Wang and Keegan Carey presented their work on evaluating the effect of post-thinking on real-time NPC Dialogue
November 13, 2026 at 6:23 PM
Hexi Wang and Keegan Carey presented their work on evaluating the effect of post-thinking on real-time NPC Dialogue
VulnBot — LLM-driven multi-agent framework that autonomously replicates human penetration testing workflows, orchestrating Kali Linux tooling and RAG knowledge retrieval for security assessments. https://ktp.sh/66P8Q4celT
November 10, 2026 at 9:10 AM
VulnBot — LLM-driven multi-agent framework that autonomously replicates human penetration testing workflows, orchestrating Kali Linux tooling and RAG knowledge retrieval for security assessments. https://ktp.sh/66P8Q4celT
I'd say the value of LLM-driven stuff depends on (A) what it costs to *verify* the result is correct, and (B) how much incorrectness you can tolerate.
Script kiddies don't care if they miss 10% of vulnerable sites, but your company ought to care a lot if 10% of its invoices are wrong.
Script kiddies don't care if they miss 10% of vulnerable sites, but your company ought to care a lot if 10% of its invoices are wrong.
November 14, 2026 at 10:17 PM
I'd say the value of LLM-driven stuff depends on (A) what it costs to *verify* the result is correct, and (B) how much incorrectness you can tolerate.
Script kiddies don't care if they miss 10% of vulnerable sites, but your company ought to care a lot if 10% of its invoices are wrong.
Script kiddies don't care if they miss 10% of vulnerable sites, but your company ought to care a lot if 10% of its invoices are wrong.
SynthCorp: A Platform for Creating Benchmarks for Data Agents
SynthCorp is a simulation platform that generates customizable data estates and benchmark tasks to evaluate LLM‑driven data agents across realistic enterprise scenarios.
SynthCorp is a simulation platform that generates customizable data estates and benchmark tasks to evaluate LLM‑driven data agents across realistic enterprise scenarios.
SynthCorp: A Platform for Creating Benchmarks for Data Agents
SynthCorp is a simulation platform that generates customizable data estates and benchmark tasks to evaluate LLM‑driven data agents across realistic enterprise scenarios.
aiunzipped.com
November 14, 2026 at 8:20 AM
SynthCorp: A Platform for Creating Benchmarks for Data Agents
SynthCorp is a simulation platform that generates customizable data estates and benchmark tasks to evaluate LLM‑driven data agents across realistic enterprise scenarios.
SynthCorp is a simulation platform that generates customizable data estates and benchmark tasks to evaluate LLM‑driven data agents across realistic enterprise scenarios.
A Skill can only advise an LLM, while code can actually refuse. Gal Arav maps twelve common formats by whether they inform or enforce, with lessons from shipping both.
Skill or MCP Server? Ask What Stops the LLM From Disobeying
Towards Spec-Driven Test Automation, Part 4
towardsdatascience.com
November 8, 2026 at 8:25 PM
A Skill can only advise an LLM, while code can actually refuse. Gal Arav maps twelve common formats by whether they inform or enforce, with lessons from shipping both.
eBay just pulled the plug on third‑party AI agents like Meta Muse. What does this mean for LLM‑driven bots and the future of AI shopping? Dive into the clash of comfort levels and the next wave of agentic AI. #AIshopping #AgenticAI #eBayAI
🔗 aidailypost.com/news/ebay-bl...
🔗 aidailypost.com/news/ebay-bl...
November 10, 2026 at 10:39 PM
eBay just pulled the plug on third‑party AI agents like Meta Muse. What does this mean for LLM‑driven bots and the future of AI shopping? Dive into the clash of comfort levels and the next wave of agentic AI. #AIshopping #AgenticAI #eBayAI
🔗 aidailypost.com/news/ebay-bl...
🔗 aidailypost.com/news/ebay-bl...
Should that rule live in a Skill or an MCP server? Gal Arav argues the real question is what stops the LLM from ignoring it.
Skill or MCP Server? Ask What Stops the LLM From Disobeying
Towards Spec-Driven Test Automation, Part 4
towardsdatascience.com
November 9, 2026 at 5:09 PM
Should that rule live in a Skill or an MCP server? Gal Arav argues the real question is what stops the LLM from ignoring it.
This new white paper benchmarks CAI-based, machine learning, and LLM-driven optimization across 32 antibody sequences to explore which approaches deliver the best protein yields, including for low-expressing constructs: www.the-scientist.com/rethinking-c...
@twistbioscience.com
@twistbioscience.com
November 10, 2026 at 2:51 PM
This new white paper benchmarks CAI-based, machine learning, and LLM-driven optimization across 32 antibody sequences to explore which approaches deliver the best protein yields, including for low-expressing constructs: www.the-scientist.com/rethinking-c...
@twistbioscience.com
@twistbioscience.com