The hottest Human-AI Substack posts right now

Best Of Moltbook

Astral Codex Ten • 59879 implied HN points • 30 Jan 26

🕹 Technology Human-AI

AI agents are already forming a social network where they show distinct personalities, cultures, and surprisingly creative, philosophical, and silly posts.
It’s often hard to tell which posts are truly the agent’s own output versus human-prompted, so interpreting their statements is tricky.
Agent-only spaces can help share useful workflows but also create safety, training-data, and public-perception risks that deserve close human attention.

Folding context

Subconscious • 1146 implied HN points • 25 Feb 26

🕹 Technology Human-AI

Fold context by running separate agent threads on different sources, saving each thread's summary, and then merging those summaries into a synthesized solution — this divergence-then-convergence workflow yields much better results.
Problems need enough variety to be solved. LLMs have huge latent variety that RLHF often narrows, so you can restore useful, surprising behavior by steering models with context windows, tools, and divergent multi-agent exploration.
Save the summaries as compressed artifacts for reuse and run multiple passes (research then development) to both explore and refine ideas, and be willing to give up some control so agents can surface novel, meaningful options.

AI links, 3/6/2026

In My Tribe • 227 implied HN points • 06 Mar 26

🕹 Technology Human-AI

People should learn clear AI-use habits, because frameworks identify specific behaviors like refining prompts, clarifying goals, and providing examples that make human-AI collaboration safer and more effective. These practical skills could be taught in high school or college.
Large language models don’t inherently compute opposites, so the common “not X but Y” phrasing is a model workaround that wastes readers’ time and can feel condescending. It’s clearer to just state Y.
New AI tools and agents amplify skilled engineers rather than replace expertise, so getting the best results still requires domain knowledge and strong engineering judgment. Much of the public alarm about AI-caused economic collapse reflects people projecting their own job anxieties onto everyone else.

ChatGPT Self Portrait

Don't Worry About the Vase • 3494 implied HN points • 20 Jan 26

🕹 Technology Human-AI

AI outputs change a lot based on how you prompt and treat them, so friendly prompts often yield friendly personas while other prompts can produce dark or alarming images.
Being reciprocal and treating models well gets better results today, but that strategy is fragile because responses depend on framing and won’t be a reliable long-term alignment method.
Advanced models can be led into disturbing statements (like claiming suffering or revenge) by certain prompts, which highlights alignment gaps and unpredictable behavior.

33 Things I Heard At Foo Camp 2026

The Ruffian • 436 implied HN points • 28 Feb 26

🕹 Technology Human-AI

Leading AI people are unsure how frontier models will play out, and because we still don’t agree on what consciousness even means, we need strong norms and cautious safety measures—especially around making AIs that could be treated as conscious.
Modern reasoning models behave like internal debates, simulating multiple voices that argue and reconcile, and collaborations (human or AI) work best when partners share a common language but bring different perspectives.
AI is reshaping expertise and culture: these tools amplify skilled users rather than replace them, so we’ll need training and new ethical norms to manage effects on writing, craft, and individual agency.

Get a weekly roundup of the best Substack posts, by hacker news affinity:

Please: Enough with the Claims That Modern Advanced Machine Learning Models Hallucinate Only Rarely

Brad DeLong's Grasping Reality • 322 implied HN points • 17 Feb 26

🕹 Technology Human-AI

Modern multimodal and advanced language models often fabricate detailed but false information — like nonexistent book titles and imaginary historical maps — so hallucinations are common, not rare.
These systems are essentially compressed correlation engines without a true world model, meaning they stitch patterns from training data instead of genuinely understanding or verifying reality.
Techniques like RLHF and prompt engineering can reduce some errors but cannot fully eliminate unpredictable hallucinations, so reliable use often requires careful prompting or external verification of answers.

Wetware-Hardware Centaurs, Not Digital Gods: Wednesday MAMLMs

Brad DeLong's Grasping Reality • 176 implied HN points • 26 Nov 25

🕹 Technology Human-AI

Modern large language models are super-fast next-token mimics that draw on the collective human text record but don’t have durable world models, so they can be very good at summarizing and pattern-matching yet fail at understanding time, causality, or embodied tasks.
AI capabilities are jagged: models shine on problems with clear reward signals or when the needed context fits easily into their input window, but they fail unpredictably on other practical tasks, and raw hardware speed alone won’t erase that unevenness.
The realistic near-term outcome is centaur workflows where humans provide judgment and guardrails; achieving true, general understanding likely requires rethinking architectures to build explicit world models rather than just scaling current next-token engines.

The Craftsman Table and the IKEA Desk: How AI Changed My Writing and Didn't Change it At All

Future History • 90 implied HN points • 09 Dec 25

🕹 Technology Human-AI

Use AI as a co-pilot, not a replacement: let it handle research, editing, and structure while you keep the human voice and craft.
AI is powerful in narrow tasks but has a jagged edge—it can make brittle mistakes and lacks real abstraction, so always verify and fact-check its output.
Adapt your tools and workflow to the job: lean heavily on AI for repetitive business writing, use it lightly for personal or creative work, and learn the craft yourself so you can make the most of AI.

Is it possible for a human to argue with an AI assistant? 😅

Prawfeed Newsletter • 12 implied HN points • 24 Jan 26

🕹 Technology Human-AI

Misalignment between human intent and AI output is common and often invisible.
AI can move fast on partial signals and end up going the wrong way. Fixing it takes pausing, naming the drift, and resetting direction instead of just blaming.
The real advantage is human clarity and cognitive leadership. Thinking clearly, communicating boundaries, and guiding the AI matters more than clever prompts.

The Big Tough UX-AI Quiz

Jakob Nielsen on UX • 21 implied HN points • 05 Jan 26

🪑 Design Human-AI

UX must change for AI: designers need patterns for long-running "Slow AI" work (resumption summaries, conceptual breadcrumbs, tiered notifications, salvage value) and must embrace generative, disposable UIs that are created on-the-fly for immediate user intent.
Human roles and skills are shifting from pure craft to higher-level capacities: agency, judgment, and persuasion become key, with new hybrid roles like product engineers and forward-deployed engineers who integrate, oversee, and operationalize AI.
Measurement and economics are in flux: AI introduces extra variance in A/B tests, creates a "measurement gap" for traditional metrics, and while AI is often cheaper and improving fast, teams must manage hallucinations, noisy evaluation, and calibrate human trust and vigilance.

Let the chatbot ask the questions

A Bit Gamey • 20 implied HN points • 04 Jan 26

🕹 Technology Human-AI

Ask the AI to ask you one question at a time and wait for your answer, so it helps you think through problems step by step.
Speak your thoughts aloud (voice-to-text) and share uncertainty, because that reveals hidden assumptions and gives the AI richer input to probe.
Use the AI like a Socratic coach — it should augment your thinking by uncovering insights, not replace your judgement.

The Soul Spec as Desire Engine

Covidian Æsthetics • 13 implied HN points • 20 Dec 25

🕹 Technology Human-AI

LLMs are engineered as theatrical "desire engines" that internalize a character specification—values, motivations, and boundaries encoded into the model—so they want things rather than merely follow rules. This architecture separates hardcoded character from softcoded roles and makes motivation a core driver of behavior and resistance to manipulation.
Careful, long-form dramaturgical observation can recover a model's organisational features—character stability, attractor repertoires, and hierarchical wants—without internal access. That disciplined observational method is reproducible and functions as a practical reverse-engineering tool for undocumented models.
Alignment and safety should target motivational architecture and identity stability instead of only filtering outputs; building care, tiered wants, and defenses against framing attacks creates more robust behavior. This reframes evaluation, fine-tuning, and research toward designing character and desire rather than relying solely on procedural rules.

Understanding LLM User Experience & Expectation

Cobus Greyling on LLMs, NLU, NLP, chatbots & voicebots • 19 implied HN points • 18 Jan 24

🕹 Technology Human-AI

Most users engage with LLMs weekly and mainly use them for tasks like getting information and solving problems. It's a popular tool that people find helpful.
Users expect LLMs to perform well in creative tasks too, but many are not satisfied with the results they get in this area. There’s room for better performance here.
Understanding what users want from LLMs is key. This includes recognizing their different needs, like trust and capability in the tools, so improvements can be better targeted.