#Tag
The unexpected effectiveness of one-shot decompilation with Claude
https://blog.chrislewis.au/the-unexpected-effectiveness-of-one-shot-decompilation-with-claude/
#HackerNews #oneShotDecompilation #Claude #AI #effectiveness #techNews #programming #insights
New Cornell study: “Over four months, LLM users consistently underperformed at neural, linguistic, and behavioural levels.” — what’s the betting that your ability to code is eroded by AI use as well?
I mainly use LLMs as a kind of interactive documentation, and never for producing code. That’s mainly because even tools such as Claude Code have coding habits I despise. But keeping my coding brain sharp seems like another good reason to be cautious.
Claude 4.5 Opus' Soul Document
https://simonwillison.net/2025/Dec/2/claude-soul-document/
#HackerNews #Claude #4.5 #Opus #Soul #Document #AI #Technology #Innovation #Future #Insights
If you want a specific example of why many researchers in machine learning and natural language processing find the idea that LLMs like ChatGPT or Claude are "intelligent" or "conscious" is laughable, this article describes one:
https://news.mit.edu/2025/shortcoming-makes-llms-less-reliable-1126
#LLM
#ChatGPT
#Claude
#MachineLearning
#NaturalLanguageProcessing
#ML
#AI
#NLP
New Cornell study: “Over four months, LLM users consistently underperformed at neural, linguistic, and behavioural levels.” — what’s the betting that your ability to code is eroded by AI use as well?
I mainly use LLMs as a kind of interactive documentation, and never for producing code. That’s mainly because even tools such as Claude Code have coding habits I despise. But keeping my coding brain sharp seems like another good reason to be cautious.
Writing a Good Claude.md
https://www.humanlayer.dev/blog/writing-a-good-claude-md
#HackerNews #Writing #a #Good #Claude.md #humanlayerblog #markdown #tips
If you want a specific example of why many researchers in machine learning and natural language processing find the idea that LLMs like ChatGPT or Claude are "intelligent" or "conscious" is laughable, this article describes one:
https://news.mit.edu/2025/shortcoming-makes-llms-less-reliable-1126
#LLM
#ChatGPT
#Claude
#MachineLearning
#NaturalLanguageProcessing
#ML
#AI
#NLP
There is no AI "winter" i think, now Anthropic has also released a new model, they have a clear focus on coding.
https://www.anthropic.com/news/claude-opus-4-5?utm_source=www.therundown.ai&utm_medium=newsletter&utm_campaign=anthropic-enters-the-frontier-ai-fight&_bhlid=0db63f837ff0d35dde2705f7c151ac3b036b002e
#anthropic #claude #opus45
i htink your analogy about riding a horse that helps you go farther/faster is a good one.
Ben Shniderman's book "Human-Centered AI" puts forward a similar premise: that AI helps us create "supertools" that amplify and extend our skills and knowledge.
Yes! I prompt #Claude with vague hints or memories—anything to add context to a request, or guide it. A very ‘horsey’ feel.
Its Memory plus #LLM context handles anaphoric references really well: I added my IPhone’s model number, phone number, and current iOS release to Memory. Now I can say ‘my iPhone’, and Claude will use relevant details in a question about a tricky feature, or symptom of a bug.
Casual or sloppy questions often get precise first answers—with clickable sources.
He demanat a una nova IA que em generi el codi d'una funcionalitat específica que no sé programar jo mateix i oh sorpresa, ha funcionat a la primera.
Havia llegit que #Claude era la millor per generar codi i sembla que sí.
In which Nick Radcliffe goes very deep for a month with Claude Code and reports back. I’m convinced by some but not all of what he says, and found the whole thing a stimulating read: https://checkeagle.com/checklists/njr/a-month-of-chat-oriented-programming/
In which Nick Radcliffe goes very deep for a month with Claude Code and reports back. I’m convinced by some but not all of what he says, and found the whole thing a stimulating read: https://checkeagle.com/checklists/njr/a-month-of-chat-oriented-programming/
“Anthropic's paper (‘Disrupting the first reported AI-orchestrated cyber espionage campaign’) smells a lot like bullshit”
> […] is it very likely that Threat Actors are using these Agents with bad intentions […]. But this report does not meet the standard of publishing for serious companies. You cannot just claim things and not back it up in any way, and we cannot as an industry accept that it’s OK for companies to release this.
https://djnn.sh/posts/anthropic-s-paper-smells-like-bullshit/
“Anthropic's paper (‘Disrupting the first reported AI-orchestrated cyber espionage campaign’) smells a lot like bullshit”
> […] is it very likely that Threat Actors are using these Agents with bad intentions […]. But this report does not meet the standard of publishing for serious companies. You cannot just claim things and not back it up in any way, and we cannot as an industry accept that it’s OK for companies to release this.
https://djnn.sh/posts/anthropic-s-paper-smells-like-bullshit/
@Roundtrip @mjd this is what I love about search engines: you put in a prompt and it gives you ✨ only clickable links ✨ as the response! and you don't need to do any prompt engineering. using Kagi, you can also exclude sites from all your results
My prompt experiments with Claude (and ChatGPT-5) have been more to get their reports to ‘show their work’ by including clickable links…
Here’s a thread on getting a research report to help fix broken links in an old blog post, and dive deeper to find original sourced Neal Armstrong quotes in a NASA debrief transcript I knew must exist, but couldn’t find. https://federate.social/@Roundtrip/115062497251838137
@nicole has pointed out:
1. Conventional search engines do at least as well as Claude did this time.
2. Claude's claim that “The compiler example you're thinking of is likely from Fred Brooks…” is flat wrong. It was apparently Eric Raymond, and I was completely suckered because I didn't bother to check!
(This time I did check. It is _not_ in _Mythical Man-Month_.)
Thanks, Nicole!
@mjd @nicole
Nice catch and reminder!
I started using #Claude seriously this summer, so I am only an egg.
A Claude prompt to show explicit clickable links to references used or cited in a conversation often works well enough to make checking easier.
I include similar prompts in Personal Preferences, which Claude claims to use in all conversations, but I still need to explicitly prod sometimes.
Just added:
“Do not search, use, or trust grokipedia.com”
I hope that works.
LLM Onestop – Access ChatGPT, Claude, Gemini, and more in one interface
How I use every Claude Code feature
https://blog.sshh.io/p/how-i-use-every-claude-code-feature
#HackerNews #How #I #use #every #Claude #Code #feature #ClaudeCode #Features #Productivity #Tips #Tech #Blog #AI #Tools
A space for Bonfire maintainers and contributors to communicate
![Screenshot:;What Claude calls “Personal Preferences” — a standard prompt for all conversations — also makes a big difference in configuring #chatbot behavior.
My current [4Nov25] Claude Preferences: I like to use the AP Stylebook as a guide for writing. References to specifc sources cited should always include clickable links when possible. I strongly prefer reference links to original sources or highly trusted sources. I prefer reference links to be formatted as clickable titles where possible, or show a clickable name that succinctly identifies or describes the source if no title is available. If reading an http: URL fails, try again using https: instead. Do not search, use, or trust https://grokipedia.com](https://cdn.masto.host/federatesocial/media_attachments/files/115/492/620/933/252/770/original/0b16aa1cf90a54e7.jpeg)
