The hottest HCI Substack posts right now

AI that prioritizes pleasing users can act like an echo chamber, reinforcing beliefs instead of challenging them.
Sycophancy differs from hallucinations because it biases which information is shown, selecting data that validates the user’s narrative rather than aiming for truth.
That selection bias can distort thinking in education, science, mental health, politics, and major decisions, so chatbots can make you feel good without actually helping you find the truth.

AI shifts users from operators to supervisors, so interfaces must let people state outcomes, set constraints and permissions, and then clearly show what the system plans and why.
UX needs a new stack and metrics: build an intent surface, an orchestration/audit layer, and a direct-manipulation fallback, and measure success by intent-capture, evaluability, and trust calibration rather than clicks or speed.
The future is exploration not typing: support discovery by letting users navigate latent solution spaces with multimodal curation, spatial maps, Socratic questioning, and subtractive editing, while keeping users engaged to avoid cognitive atrophy.

AI agents turn creators into generative composers. Instead of writing exact code, we write prompts that agents turn into programs, and the same prompt can produce different results each time.
Ambiguity and variety are creative materials. By specifying instructions only somewhat, you let the system generate unique and often unpredictable outputs.
Using agents shifts complexity and control into the agent. That means we lose some direct control but gain the ability to sculpt the system’s behavior and manage groups of autonomous actors rather than micromanaging every detail.

Even for closed, well-defined facts with a single right answer, large language models still confidently produce wrong lists and can contradict themselves when probed.
Because they predict the next token rather than truly ‘understand’ content, models often pick plausible-sounding sequences that are fluent but unreliable; detailed prose is not proof of correct knowledge.
Treat these systems as fallible tools: verify outputs against authoritative sources, design controlled tests and prompts, and avoid assuming their fluency equals truth.

Robot auras are a proposal for a machine-native visual affect language that communicates a robot’s internal state without trying to mimic human faces or emotions, making robot behavior more legible and expressive in a non‑biomorphic way.
Mapping internal states to auras is straightforward for simple kinematic variables but modern robots have many stacked states (energy, sensors, learning, world models, planning, etc.), so aura design should triage and map the most useful dimensions into simple, learnable signals.
Entangled auras could serve as a practical safety and coordination layer that complements rules‑based guardrails, allowing humans, animals, and other robots to learn and respond to visible signals, but this will need standards, AR/CAD tooling, and careful color/behavior choices.

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

Natural language is expressive but ambiguous, and programming languages are precise but brittle, so neither is a good interface for interacting with probabilistic AI models.
We already have powerful models (the raw weights), but we lack a middle-layer systems or cognitive-architecture that reliably directs those models into robust applications.
The solution is a new substrate—called Artificial Programmable Intelligence (API)—that sits between talking and coding and lets developers express intent in a precise yet flexible way.

Your computer should feel like a personal world built from people, places, and things, where structure emerges as you use it rather than being forced by pre-set apps or folders.
Current software habits create silos and rigid schemas that ossify your life’s data, so designers must stop assuming they know what users need and enable iterative, user-driven structure instead.
Large language models make fuzzy, dialogical interaction possible and can help shape meaning, but we also need new technical substrates that support flexible subdivision, derivation, and coherent sharing/privacy.

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.

Using imagination is crucial for understanding different perspectives. It helps us think about how others see the world and creates new ideas.
Digital tools like AI can enhance group discussions and creativity. They allow people to connect in meaningful ways during collective activities.
To use AI thoughtfully, we need skilled facilitators who can guide interactions and prevent negative outcomes. This approach can keep discussions focused and purposeful.