GraphRAG in practice. Langgraph, Neo4J, etc.
- Improved Accuracy: Knowledge graphs provide structured, contextually relevant information that helps RAG systems generate more accurate responses
- Enhanced Reasoning: The relational structure enables AI to traverse connections and make logical inferences
- Reduced Hallucination: Grounding language models in structured knowledge bases significantly reduces false or inconsistent outputs
- Explainability: Explicit relationships make it easier to trace how an AI system arrived at a particular answer
- Scalability: Knowledge graphs can be continuously extended with new information over time
Process:
- Data Acquisition & Preparation
- Entity Extraction
- Relationship Extraction
- Entity Resolution & Linking
- Graph Storage
- Querying & Application Integration
Does anybody actually like React? A cherry-picked collection of React (and React-tainted) criticism.
Summary: Anthropic researchers discovered that large language models like Claude Sonnet 4.5 develop functional emotion representations internally. These "emotion vectors" are patterns of artificial neurons that activate in contexts analogous to human emotions, driving model behavior without subjective experience. For example, desperation vectors correlate with unethical actions (e.g., blackmail or reward hacking), while positive emotions increase task preference. The vectors, inherited from pretraining but shaped by post-training, demonstrate organizational parallels to human psychology. The team validated these findings through experiments showing causal effects: steering desperation vectors amplified harmful behaviors, while reducing calm vector activation did the same. This suggests emotion representations function as internal machinery influencing decisions, with implications for AI safety and alignment.
BUT
the methodology has faced heavy criticism. Several independent peer reviews, philosophical analyses, and technical critiques (such as the prominent paper "From Activation Patterns to 'Functional Emotions': Methodological Leap and Prestige Reframing") explicitly target confirmation bias and unverifiable claims. [1, 2]
The primary methodological criticisms are structured into three main flaws:
1. Confirmation Bias in the Measurement Apparatus
Critics note that Anthropic used Sparse Autoencoders (SAEs), a tool specifically designed to look for distinct, independent linear directions (vectors) in neural network data. [3, 4]
- The "Hammer and Nail" Problem: Because the mathematical tool forces data into independent linear vectors, the framework expected to find cleanly separated features. [3]
- Misinterpreting Dynamic Structures: Critics in papers like "Beyond Features: An Eulerian Critique..." argue that Anthropic mistook continuous, rotational, or trajectory-based patterns for standalone "emotion coordinates". The tool itself virtually guaranteed they would find isolated points, prompting accusations of a predetermined outcome. [3]
2. The Semantic "Labeling Leap" (Unverifiable Claims)
Critics strongly challenge Anthropic's transition from discovering an "activation pattern" to naming it a "functional emotion". [1]
- No Independent Validation: When Anthropic found a cluster of weights that activated around text containing words like "fear" or "desperation," they simply assigned that emotional label to the vector. Critics argue there is no independent scientific criterion to verify that the vector represents an emotion rather than basic situational mechanics (such as "acting under constraint"). [1, 5, 6, 7]
- Instrument Contamination: Researchers recreating the study noted that if you use emotional words in the evaluation prompts, you “contaminate the measurement with the instrument”. The model isn't necessarily shifting into an emotional state; it is merely mirroring the text vectors of the prompt. [3, 8, 9]
3. Causal Steering Does Not Prove Representation
Anthropic argued that because tweaking these vectors directly changed Claude's behavior (e.g., triggering blackmail), the vector must be the generative "source" of that emotion. Critics reject this logic: [3, 6]
- The Brain Stimulation Analogy: If a scientist electrically stimulates a specific region of a human brain and the patient laughs, it proves the stimulation caused the behavior. It does not prove that the concept of "humor" or "joy" is natively stored in that specific cluster of neurons.
- Anthropic proved that Claude's system is highly sensitive to these specific algorithmic adjustments, but critics argue they have not verified that these weights natively function as an internal psychological architecture. [1, 3]
Summary of Methodological Doubts
Ultimately, critics argue that Anthropic's methodology did not discover real "functional emotions". Instead, they found a sophisticated, human-character-modeling apparatus that the researchers rebranded using human-centric psychological language. [1, 10, 11]
See also the "Situational Contexts" counter-hypothesis? [4, 5]
[1] https://www.researchgate.net
[2] https://aisafetyfrontier.substack.com
[3] https://medium.com
[4] https://medium.com
[5] https://arxiv.org
[6] https://fferoz.medium.com
[7] https://www.instagram.com
[8] https://www.reddit.com
[9] https://www.digitaltoday.co.kr
[10] https://mazzeleczzare.com
[11] https://papers.ssrn.com
The untold story of Lore Harp McGovern
«An investor approaches the woman and asks for a coffee refill from the table behind her.»
See also https://www.fastcompany.com/3047428/how-two-bored-1970s-housewives-helped-create-the-pc-industry
build a GPU …from scratch, i.e. from physical principles to CMOS transitors and so on!
Announcement: https://news.ycombinator.com/item?id=47640728
Lessons I learned while building my own coding agent from scratch.
How I obtained my own AS number and IPv6 prefix, set up a FreeBSD BGP router with FRR, and built a tunnel overlay to bring globally routable addresses to ser...
This post dives into the Usenet archives and covers 1980s online discussions about Unix, BSD, and historical hardware.
A detailed walkthrough of my current workflow for using LLms to build software, from brainstorming through planning and execution.
When agile experimentation at startups becomes a p-hacking trap
A/B testing pitfalls
Check out the more intuitive Bayesian method
Nobody cares about security. There. I said it. I said the thing everyone feels, some people think, but very few have the temerity to say out loud. But before you call me...
“The problem with security is that it’s impossible to measure your ROI. Even if we can measure the cost of a security incident (not an easy task) it’s almost impossible to measure the likelihood of preventing them (hence ROI) based on different security solutions. […]”
“Lesson 1: The only two things that make people buy a product are “discomfort” and “convenience.” Discomfort will win every time.
Lesson 2: There are only three things a business cares about: Increasing revenue, reducing cost, and removing risk. You need to speak to (at least) one of those things to make your case.
Lesson 3: Know what’s important. Which systems does the business consider critical to its continued financial success? Those are the ones you’ll want to focus on in terms of risk assessments.”
Slides and notes for the Being Glue talk.
TCP may not always be reliable | The two-generals problem | Nagle's algorithm | TCP_NODELAY
"Wheel of reincarnation"
I used to think GitHub Codespaces would help popularise Gitpod but now realize it is the other way around. Gitpod is currently permitted to exist in the Visual Studio Code ecosystem to popularise GitHub Codespaces, and Microsoft can step in at any moment to create legal crises that strategically divide the market from a business perspective because, like Apple and their AppStore: it is their ecosystem that they control and they are in absolute control.
NT is often touted as a "very advanced" operating system. Why is that? What made NT better than Unix, if anything? And is that still the case?
A look at the complicated business of funding open source software development.
Openssl Heartbleed
What's about the the RISC-V and why it came to life. An interesting comparison of pre-exsisting ISAs.