Raghav Agrawal @impactology@mastodon.social

@impactology

Former Diamond Dealer | Turning research on learning, cognition, and systems into interfaces that reduce confusion, cost and decision stress.

3.533 Folgt2.288 Follower6.070 Posts

X X YouTube YouTube Bluesky

Ist das dein Profil?Übernimm es — dann gehört dir impactology.dirty.social.Profil übernehmen

Raghav Agrawal @impactology@mastodon.social@impactology· 11d

The "ability to" statements is inspired by Jon Kolko's product management course www.udemy.com/course/produ... and the software abilities part is inspired by Peter Denning's principles of computing x.com/impactology/...

ABILITY TO STATEMENTS slide.
Heading: 'A capability is framed as an ability to statement.'

Text explains: By structuring capabilities as succinct 'ability to' statements, we shift from thinking about what the product has to what the product helps someone do.

An 'ability to' statement: describes what someone can do with a product (includes a verb), ignores technology and focuses on human wants/needs/desires, doesn't describe how someone does something, leaving mechanism for action to be defined later.

ABILITY TO STATEMENTS slide titled 'Identify the capabilities.'

Lists four questions: 1) What behaviors need to change? 2) What emotions need to change? 3) What actions need to take place? 4) What attitudes need to adjust?

Green sticky note reads: 'A tool that helps students connect with alumni to get feedback on school work and gain tutoring and mentorship.'

User goals listed: student wants help with schoolwork, gain confidence in choices, wants supportive mentor; alumni wants to give back to university, wants to act as mentor."

ABILITY TO STATEMENTS slide titled 'Identify the capabilities.'

Shows green sticky note about student-alumni connection tool and five user goals.

Right side shows Capabilities column with five gray boxes: '
The ability to find an alumni that matches their interests and passions,'
'The ability to reach out to an alumni and ask for help,'
'The ability to explain the types of help they need,' '
The ability to volunteer to be a mentor,'
'The ability to learn how to be an effective mentor.'

Raghav Agrawal @impactology@mastodon.social@impactology· 12d

For translating a user need/concern/frustration/breakdown/research insight into a design concept ask What would the person need to become able to do here? and then What must the computer be able to do so the person gains this ability?"

Two-column framework showing steps 1 and 2 of a human-centered software design process.

Left column (Step 1): 'Observe the context' - Instructs to start with recurring situations: home, neighbourhood, work, education, small business, healthcare, public services, research. Also instructs to capture interruptions, workarounds, repeated decisions, information loss, uncertainty and consequences. Note at bottom: 'The first output is a Situation Record.'

Right column (Step 2): 'Frame the activity' - Instructs to convert observations into actors, goals, objects, actions, sequences, decisions, dependencies, breakdowns, exceptions and consequences. States that user research becomes an operational model, not a report. Note at bottom: 'Construct only as much context as the task requires.

Two-column framework showing steps 3 and 4 of a human-centered software design process.

Left column (Step 3): 'Translate into abilities' - Instructs to ask for every breakdown: 'What would the person need to become able to do here?' Lists human abilities: Remember what happened, Compare alternatives, Detect exceptions, Coordinate, Verify, Return to unfinished work, Reverse a mistake. Note at bottom: 'These become Human Ability Statements.'

Right column (Step 4): 'Derive the software capability' - Instructs to ask: 'What must the computer be able to do so the person gains this ability?' Lists software capabilities: Persist state, Reconstruct history, Classify events, Extract intent, Normalize data, Request confirmation, Preserve reversible history. Note at bottom: 'Only now do APIs and libraries enter.'

The framework emphasizes translating human needs into abilities first, then deriving technical requirements, before considering implementation tools.

Raghav Agrawal @impactology@mastodon.social@impactology· 17d

Below all of this, add four layers. 1. What were you hoping to do 2. The UI primitives used 4. Software capability and the building blocks used (APIs, libraries, algorithms)

A diagram template for mapping user interface flows across three sequential steps. Each step shows "What I see" above a line and "What I do" below it. Underneath, rows track the human goal, interface job, UI primitives, software ability, and technical building blocks for each step.

Raghav Agrawal @impactology@mastodon.social@impactology· 19d

One thing that I want to do is build is a reference for myself that describes a bunch of apis, libraries with their human use case, like their jtbd. Same for UI primitives. So that you can hear a vague idea & be able to practice sketching out the entire end to end flow & discuss technical trade-offs

A Bridge Between Human Abilities, Software Capabilities and Interface Design' describing the Capability Atlas.

Purpose: To build a practical translation layer between human needs, software capabilities, and interface design, bridging designers' understanding of user goals with engineers' technical knowledge through a shared vocabulary.

Core Principle: Every technology gives people a new ability, described as simple 'Ability Statements' rather than technical language.

Examples include: remember information, compare large amounts of data, understand speech, generate text, verify identity, detect location, synchronize devices, notify people at the right moment, search large collections, and automate repetitive work.

Raghav Agrawal @impactology@mastodon.social@impactology· 20d

All tech jobs should be like this razorpay.com/ai-builders/ "No resume theatre. Just show us what you've built.' Three steps. 1. Fill the form, 2. Submit your project or GitHub, 3. If it has signal, we'll call in 48 hrs" x.com/harshilmathu...

Tweet from Harshil Mathur (@harshilmathur) announcing an always-on opening at Razorpay for AI builders. The tweet praises the design team's career page at razorpay.com/ai-builders/ and includes a video thumbnail showing the page's dark interface with text about building the future together and values like '/UNLIMITED TOKENS' '/ANY MODEL' '/AI TOOL' '/NO HIERARCHY' '/SHIP FAST'

Razorpay AI builders career page hero section with dark background showing an iMac computer with code on screen. Headline reads: 'No resume theatre. Just show us what you've built.' Below: 'Three steps. No nonsense.' followed by: 1/ Fill the form, 2/ Submit your project or GitHub, 3/ If it has signal, we'll call in 48 hrs. Blue 'Apply Now' button visible.

Razorpay AI builders page section with dark code-patterned background. Heading: 'Unleash Your Inner Builder.' Text explains: 'We aren't looking for standard resumes, we want your best AI build that solves a real, tough problem. Show us how you prioritize features to build something original. Use any tech stack you want, just make sure it's fresh work built for this challenge that prioritizes core logic over surface polish.' White 'Start' button below.

Razorpay AI builders application form step with dark background. Heading with sparkle emoji: 'Show us the magic.' Instructions: 'Paste a live link or working demo. Deployed somewhere we can click and try. (Vercel, Replit, HuggingFace Spaces, your own domain, whatever)' Additional text: 'If it can't be deployed, drop a link to a 2-minute screen recording of it working instead.' URL input field showing 'https://' placeholder and 'Next' button below.

Raghav Agrawal @impactology@mastodon.social@impactology· 21d

Still trying to figure out where and how should the interaction for writing the context for the links appear, should it be a pop up modal like a bookmark manager or something else

A screenshot of a browser's "Edit bookmark" dialog box.

Raghav Agrawal @impactology@mastodon.social@impactology· 21d

Also trying to build this

Inquiry Atlas is a link-capture tool for people who research obsessively and lose the thread.

A browser extension lets you grab open tabs or pull from a saved link archive. Multi-select a cluster, and immediately attach the question you were pursuing and the tags that make it retrievable. That grouped cluster, with its context intact, lands in a shared workspace where your entire link collection lives as an organizable corpus.

The workspace is a fast card-sorting interface: select across hundreds of links, group them, name the inquiry, move on. No Kanban, no workflow states, just a flat pile becoming an ontology, as fast as your hands can work.

The output, an interactive site walking through your research collection as a set of named inquiries is where it's going. But the problem it solves first is the moment of capture: the question that led you somewhere, attached to the links before you forget why you opened them.

A capture-first, sort-fast, publish-beautifully system for people who open 100 tabs and forget the question that started it all.

Raghav Agrawal @impactology@mastodon.social@impactology· 22d

He built a knowledge graph of interface patterns, their principles, history, and similar examples. When he documents a new UI concept he uses a command to search thousands of related patterns and principles, matching them to the current concept

Network diagram showing interconnected interface design patterns and principles. Nodes include labeled boxes for principles like "Streaming converts waiting into reading," "Decomposition is generative," and "Reverse every action," connected via colored lines to pattern examples like "StretchText," "ChatGPT Canvas," "Expand and Collapse," and "EuterPen." A photo of Jocelyn Shen appears on the left. The graph visually traces relationships between design concepts, historical examples, and underlying principles.

Focused view of the knowledge graph highlighting the "Expand-in-Place Disclosure" pattern node in the center, with teal connection lines linking it to related concepts including "Writing with Clay," "StretchText," "Telescopic Text," "Nutshell," and "Expandable Inline." Other nodes fade into the background, emphasizing how one pattern connects to multiple principles and historical examples.

Detailed documentation page titled "EXPAND-IN-PLACE DISCLOSURE" showing visual examples of StretchText, Telescopic Text, and Nutshell interfaces. Below, text explains alternative names (stretchtext, inline expansion, expandable inline explanation) and describes the problem: balancing concision against thoroughness when writing for both experts and newcomers. The page demonstrates how the knowledge graph connects to detailed written analysis.

Raghav Agrawal @impactology@mastodon.social@impactology· 22d

interfacelineage.com "Interface Lineage takes the invisible mechanisms in your interfaces apart for the people who make them. You leave able to name each move, feel how it works, and say why it matters" A library of interaction mechanisms and how they came about

Checkmark list showing what's included in Interface Lineage: hundreds of interfaces analyzed to the mechanism level, principles they prove or break, reusable patterns with origins, maker context, interactive demos, optional quiz cards and reveals, and a critical analytical approach throughout.

Six Thumbnail Cards:
1. Press-and-Release Menus
iPhone screen showing a press-and-hold context menu appearing over a list of contacts, demonstrating long-press interaction pattern

2. Infinite Scroll Timeline
Phone interface showing an infinite scroll timeline with stacked cards or messages, demonstrating continuous vertical scrolling behavior

3. Expand and Collapse
Stacked layered cards or panels in an accordion-style interface, showing expand and collapse mechanism with visible depth and layering

4. Cursor Affinity in Rich Text Editors
Text editor interface showing formatting options (Bold, Italic, Code, Highlight) with cursor positioned at the boundary between formatted and plain text, demonstrating cursor affinity behavior

5. Clear App's Gesture-Only List
Hand holding iPhone performing a swipe gesture on an orange-colored interface, demonstrating gesture-based list item deletion or clearing interaction

6. Sketch to Search
Ceramic bowl or vessel photographed on white surface, representing sketch-to-search functionality where visual input finds similar items

Raghav Agrawal @impactology@mastodon.social@impactology· 29d

"Juniors embrace AI as an always-available first mentor to reduce dependency on busy seniors" "J1 delegates simpler questions to AI, such as “how do I find this in a file [or] what does this function do?”

Text excerpt describing how junior engineers use AI as a mentor. The opening statement in bold italics reads: 'Juniors embrace AI as an always-available first mentor to reduce dependency on busy seniors.' The text explains that juniors used AI for basic questions requiring little to no context. Participant J2 likened AI chats to 'talking to...a smart person that knows everything' and admitted pinging teammates less. J9 stated that senior engineers are often too busy for thorough code review, and explained that while Agentic AI reduces how often they need to ask questions, it changes expectations about seeking help from seniors. J9 recalled their mentor implicitly suggesting 'there's an answer for everything on the Internet,' interpreting this as 'why are you asking me so many questions?' J1 delegates simpler questions to AI, such as 'how do I find this in a file [or] what does this function do?' while J8 states their job requires finance knowledge not learned in school, making AI very helpful for teaching them

Raghav Agrawal @impactology@mastodon.social@impactology· 29d

From Junior to Senior: Allocating Agency and Navigating Professional Growth in Agentic AI-Mediated Software Engineering by Dana Feng, Bhada Yun, April Yi Wang arxiv.org/abs/2602.00496

Three tables displaying participant demographics across three study phases. Table 1 shows Phase 1 participants (S1-S5): 5 senior software engineers with 5-13 years experience from companies including Big Tech/Cloud & AI, FinTech, Design Tools, DevTools startup, and Enterprise SaaS, detailing their GenAI and Agentic AI tool usage and familiarity levels. Table 2 shows Phase 2 participants (J1-J10): 10 junior software engineers with less than 1 year to 1 year experience from various sectors including FinServ startup, FinTech, Enterprise SaaS, HealthTech, Big Tech, FinMedia, Investment, Payments, and Consulting, all using Software Engineer titles with varying GenAI and Agentic AI tool usage. Table 3 shows Phase 3 participants (S6-S10): 5 senior engineers with 8-26 years total experience (6-16 years senior level) from FinTech, Social/Consumer, Big Tech/Cloud & AI, and Insurance/Health sectors, in roles ranging from Lead to Staff to Senior Software Engineers, with their AI tool usage patterns.

Text excerpt discussing practical challenges in implementing Prompt Code Reviews (PCRs). The text explains that code review is time-intensive and reviewing full prompt histories could overwhelm reviewers. Two sections are highlighted in yellow: First highlight states 'Juniors could self-curate 2-3 key prompts that shaped their solution, while automation could filter out sensitive or exploratory content.' Second highlight states 'By making AI interactions part of the review process, teams can preserve the human judgment and ownership that both juniors and seniors identified as essential to meaningful work.' The text concludes by noting that seniors reviewing junior artifacts in brief 12-minute sessions identified patterns of over-reliance, suggested better prompting strategies, and caught instances where juniors accepted incorrect AI suggestions.

Two tables displaying prompt types and frequencies used during Phase 2. Table 4 shows Agent mode prompts: Debugging (24 instances, highest frequency) with sample prompt about app stalling on Loading Activity logs; Code explanation (8 instances); Command (4 instances); Command/Code explanation (2 instances); Command/Debugging (2 instances); Concept Explanation (2 instances); and Asking for ideas (1 instance). Table 5 shows Ask mode prompts: Code explanation (12 instances, highest frequency) with sample prompt 'What is the logs-container div doing?'; Debugging (10 instances); Concept Explanation (3 instances); Command (1 instance); and Verification (1 instance). The tables contrast Agent mode's debugging-focused usage with Ask mode's code explanation-focused usage patterns.

Raghav Agrawal @impactology@mastodon.social@impactology· May 27

ClimateTriage - Impactful open source contributions climatetriage.com Launch your open source journey in climate and sustainability. Explore current Good First Issues from open source initiatives focused on protecting natural resources. @opensustaintech.bsky.social

ClimateTriage website interface showing a platform for finding open source climate and sustainability projects. Left sidebar displays 'Browse by Category' filters including Biosphere (43), Climate Change (41), Energy Systems (40), and other environmental topics, plus 'Browse by Language' filters showing Python (141), R (25), JavaScript (10), and other programming languages. Main content area shows 'Sort By' options and a search bar, followed by five climate-focused open source project listings: ClimKern (Python, 26 stars), CMIP7 Data Request Software (Python, 22 stars), PyBOP battery modeling tools (Python, 186 stars), censusdis U.S. Census data package (Python, 131 stars), and Wild Me Scout animal census tool (JavaScript, 11 stars). Each project displays description, programming language, star count, last activity date, and number of available issues marked with blue badges.

Raghav Agrawal @impactology@mastodon.social@impactology· May 23

The Making of Indian Statistics | Alter Magazine altermag.com/articles/the... The scrolling animation is something I've never come across from an Indian editorial, so good!

Raghav Agrawal @impactology@mastodon.social@impactology· May 15

[Open Access] Learning Environments Design Reading Series pressbooks.pub/learningenvi... Learning to Solve Problems: A Handbook for Designing Problem-Solving Learning Environments by David Jonassen staibabussalamsula.ac.id/wp-content/u...

Raghav Agrawal @impactology@mastodon.social@impactology· May 15

Enhancing Science Education Exploring Knowledge Practices with Legitimation Code Theory Edited By Margaret A.L. Blackie, Hanelie Adendorff, Marnel Mouton legitimationcodetheory.com/wp-content/u...

Raghav Agrawal @impactology@mastodon.social@impactology· May 13

Who in UX is the best person to help understand the entire tooling for private LLMs? Like the UX behind each step of making unstructured data usable for on prem self-hosted private LLMs?

• Ingest: Unstructured.io + Tesseract (scanned receipts, bank PDFs)
• Clean: Presidio (redact SSN/EIN), dedupe duplicate invoices
• Classify: DistilBERT (tax doc vs. receipt vs. bank statement)
• Chunk: LlamaIndex (preserve table structure in Excel/PDF)
• Embed: BGE-large + Qdrant (with client/matter ACLs)
• Retrieve: Hybrid search + permission filters (only see your clients)
• Generate: Mistral Small 3.1 / Qwen 3.5 (Apache 2.0, on-prem)
• Observe: Langfuse (trace every calculation), Prometheus (GPU monitoring)
• Control: Keycloak (SSO) + Cerbos (client-level permissions)
• Automate: n8n (route exceptions to senior accountant)

USE CASES:
• Auto-categorize expenses from receipt photos
• Flag audit risks: missing docs, unusual variances
• Draft tax forms from structured workpapers
• Reconcile bank feeds vs. GL entries
• Generate client-ready financial summaries

COMPLIANCE:
• Data never leaves firm servers
• GDPR/IRS-ready audit logs (7-year retention)
• Client-level isolation (no cross-client data leak)
• Role-based access: staff ↔ manager ↔ partner

Raghav Agrawal @impactology@mastodon.social@impactology· May 9

Tempeh 51g of low fat vegetarian protein in a single plate

Several strips of golden-brown stir fried tempeh arranged on a white plate with a grey wavy rim pattern. This serving contains 51g of low fat vegetarian protein in a single plate.

Raghav Agrawal @impactology@mastodon.social@impactology· May 7

Okay found something Resilient Design for UX by Johanna Sneed link.springer.com/book/9798868...

About this book

Resilient Design for UX introduces a new way of thinking about user experience, service, and product design by helping practitioners anticipate failure, embrace unpredictability, and design for resilience. Unlike traditional design methodologies that assume predictable users and stable systems and the 'best case scenario,' chaos design acknowledges that failures, disruptions, and unpredictable user behavior are inevitable. The book takes inspiration from Chaos Engineering (failure injection in systems), Resilience Engineering (anticipate/respond/learn principles), and Complexity Theory (emergent user behavior and nonlinear outcomes) to help build more resilient digital experiences that can withstand real-world unpredictability.

Raghav Agrawal @impactology@mastodon.social@impactology· May 4

How to hold 100+ screen flow in working memory sketch

Raghav Agrawal @impactology@mastodon.social@impactology· May 4

This question reminded me of what @miniver.bsky.social had said long back. Remember flows by remembering the scenarios and the goals in them

Raghav Agrawal @impactology@mastodon.social@impactology· Mar 16

Another idea is for people from big companies working with outsiders to work on co-writing articles/workshop/talks together i.e both building their personal brands together. Especially if both align on key ideas, principles.

Raghav Agrawal @impactology@mastodon.social@impactology· Mar 16

Some ideas

Raghav Agrawal @impactology@mastodon.social@impactology· Mar 6

Is this common in freelancing? What can be done to avoid it? x.com/Hartdrawss/s...

Tweet by Harshil Tomar (@Hartdrawss): "This is the TOXIC Indian freelance industry. got a legal notice today. Rs. 10,00,000 penalty for 'non-compliance'. here's the full story: - signed a ₹X L ERP project - delivered 100% of agreed scope - they kept adding features saying 'this was discussed day 1' - i absorbed it, built 40% extra out of goodwill - 6 months of over-delivering - still got a legal notice. the Indian market has a specific breed of client. they sign a scope, ignore it the moment work starts, gaslight you into believing their new demands were 'always part of the deal' and then threaten you when you can't keep up. > the signed agreement means nothing to them > your goodwill becomes their weapon > I should have walked away at month 2. > lesson learned the hard way." Includes a screenshot of an email imposing a Rs. 10,00,000 penalty for non-compliance with the agreed timeline.

Email reading: "Dear Harshil, Greetings for the day. As per the subject stated above, [REDACTED] issued a work order for the development of a web application on [REDACTED]. During the initial kickoff meeting, our Managing Director explained the complete requirements of the application. As per the commitment made, the project was to be completed [REDACTED]. However, as of March 2026, the application has reached only about 10% of our company's requirements. This unacceptable delay has resulted in a significant loss of company time and operational efficiency. Due to the prolonged delay in delivering the agreed project, a penalty of Rs. 10,00,000 has already been imposed on you for non-compliance with the agreed timeline.

Reply tweet by Kanak (@kanak_k365): "Something similar happened (not the legal notice part) he kept on adding new things, i kept on agreeing and in a meeting he says you don't know what you are doing, we should have hired someone else (a lot more) when I said I'll charge more from now on as you are just adding more and more. Sent him his money back and deleted the repo." Quoted tweet is Harshil Tomar's original post about the toxic Indian freelance industry.

Raghav Agrawal @impactology@mastodon.social@impactology· Mar 3

Aka defining a boundary of relevance right?

Herbert Simon argued that we manage complexity not by knowing more, but by constraining the problem.

Compression is not summarization. Compression is removing dimensions that do not change action.

Humans cannot process infinite complexity, so we construct boundaries to make problems solvable.

When you feel mentally heavy or blank, it’s often because you’re treating every dimension of a domain as if it matters equally. You’re trying to solve an infinite problem.

Raghav Agrawal @impactology@mastodon.social@impactology· Nov 26

@meganakpeters.bsky.social "My research aims to reveal how the brain represents and uses uncertainty, and performs adaptive computations based on noisy, incomplete information" sites.google.com/site/meganak... Wow!

Who I am

I am an Associate Professor in the UCI Department of Cognitive Sciences and a part-time Lecturer in the Department of Experimental Psychology at University College London. My group will move from UC Irvine to University College London in spring/summer 2026.

I’m also an affiliate faculty member of the UCI Department of Logic & Philosophy of Science, a Fellow in the UCI Center for the Neurobiology of Learning and Memory and the UCI Center for Theoretical Behavioral Sciences, and a Fellow in the Canadian Institute for Advanced Research Brain Mind & Consciousness program. I am an elected member of the Board of Directors at the Association for the Scientific Study of Consciousness, and advisor to several corporations and nonprofits including Conscium and PRISM.

And if that isnt enough labels for you... I am also President, Co-founder, and Chairperson of the Board of Directors at Neuromatch, where we have built a scalable, accessible, and democratized educational and community-building enterprise serving 40,000+ participants (13,000+ in our interactive Academies alone) in 130+ countries globally across computational neurosciences, deep learning, computational climate science, and neuroAI.

Previously, I was on the faculty at UC Riverside in the Department of Bioengineering, with affiliations in the Department of Psychology and the Interdepartmental Graduate Program in Neuroscience. I received my Ph.D. in computational cognitive neuroscience from the UCLA Psychology Department, having worked with Ladan Shams, and then was a postdoc there working with Hakwan Lau.
What I care about

My research aims to reveal how the brain represents and uses uncertainty, and performs adaptive computations based on noisy, incomplete information.

Raghav Agrawal @impactology@mastodon.social@impactology· Oct 17

UI Design worksheet Inspired by Five Design Sheets fds-design.github.io IxD process steps by Jonathan Korman

A comprehensive one-page design worksheet template laid out in A4 format with a dark header reading 'DESIGN WORKSHEET'. The top contains fields for PROJECT, DATE, and SHEET number.

The worksheet is organized into 9 main sections arranged in a structured grid:

Top row (3 columns):
- Section 1 'BRAINSTORM' (left): Contains fields for 'Generate Ideas', 'Filter Categorize', and 'Combine Refine' with horizontal lines for writing
- Section 3 'INTERFACE SKETCH' (center): Large dashed-border box for sketching rough layout with placeholder text 'Rough layout sketch (What it looks like)'
- Section 5 'FOCUS OPERATIONS' (right): Fields for 'Central Idea/Novel Method', 'Key Operations', and 'Controls Needed'

Second row continues sections:
- Section 2 'SYSTEM MAP' (left column): Fields for 'Main objects (nouns)', 'Actions (verbs)', 'Flow: object → action → result' with a dashed sketch area, plus 'Goal of screen' and 'Key user actions'
- Section 4 'INTERACTION' (center column): Detailed fields for Element, Trigger, Rule/validation, Data used, Action/goal connection, Feedback, Loop, and Mode
- Section 6 'VISUAL LOGIC' (right column, upper): Instructions to use spacing/size/typography only, with dashed sketch area for grayscale study
- Section 7 'DESIGN REASONING' (right column, lower): Three prompts - Problem (what was unclear?), Principle (what rule/idea fixed it?), Choice (what did you decide and why?)

Raghav Agrawal @impactology@mastodon.social@impactology· Oct 13

This means designing tools that unbundle the university’s service of bounded scope, giving self-taught learners the compression, context, and collaborative rhythm of formal education without the hierarchy or exclusion.

Unbundling of a university's service of disciplined scoped learning with rigorous but playful experimentation with other people into three chunked products

1. Contours, a custom search engine client that inverts the paradigm of unlimited scrolling and instead helps in chunking people's explorations

Does this by helping identify core transformative ideas, concepts, questions of a domain to help know what to focus & to see the boundaries of where understanding something ends.

2. CALT (Curiosity aligned learning topologies)

A platform that creates and curates hard problems to solve for you based on your existing knowledge, skills, interests

Basically it creates contextualized problems for you to solve

3. CoLearn

A platform that matches mentors & mentees based on shared interests, goals, thinking styles, values to co-create lesson plans

A clean, minimalist table titled 2. How Each Tool Produces One Dimension of This State.”

It lists three learning tools — Contours, CALT, and CoLearn — across four columns: Tool, Function, Micro-Cognitive Effect, and Macro-Cognitive Outcome.

Contours: Function — builds epistemic boundaries to show what’s relevant versus noise. Micro-effect — gives a learner the sense of finitude (“I can see the map. I can stop exploring now.”). Macro-outcome — moves from overwhelm to Clarity.

CALT: Function — translates curiosity into deliberate challenges suited to capability. Micro-effect — converts passive knowing into procedural flow (“I can try this now.”). Macro-outcome — moves from paralysis to Momentum.

CoLearn: Function — creates reciprocal learning loops by matching mentors and peers. Micro-effect — makes learning socially mirrored (“My curiosity connects to others.”). Macro-outcome — moves from isolation to Belonging.

Raghav Agrawal @impactology@mastodon.social@impactology· Oct 9

What a UI design course or book should be like Purely skill-focused, execution-oriented handbook for building useful, usable, and emotionally intelligent interfaces from sketch to production.

The Full-Stack Design Engineer’s Field Manual
A skill-based, execution-oriented handbook for building useful, usable, and emotionally intelligent interfaces — from sketch to production.
I · Model the System
• Build clarity before pixels.
• Map the conceptual model — what objects exist, what they can do, and how they relate.
• Compare system, designer, and user models to catch mismatched assumptions.
• Define interface language — verbs, nouns, and visible states — that reflect the model.
• Draw one diagram that links data, interaction, and feedback.
• Prototype system logic in a quick interactive scaffold (spreadsheet, flow builder, or sandbox).
• Output → a single-page system model that can guide every design choice.
II · Structure the Interaction
• Shape flows and states that match system logic.
• Use Five Design Sheets to explore multiple structural solutions, not just visual variants.
• Define interaction units (tasks) and their entry, transformation, and exit states.
• Diagram the behavior logic: triggers, rules, feedback, and loops.
• Combine related flows into a state map that exposes dependencies and reuse.
• Output → a tested interaction backbone: every possible state accounted for.

IV · Express the Logic Visually
• Turn invisible behavior into visible structure.
• Build a visual grammar (type, color, space, iconography) tied to functional roles.
• Represent state change through clear visual transitions, not arbitrary motion.
• Balance procedural, participatory, spatial, and encyclopedic affordances (Murray).
• Design for hierarchy of attention: one dominant action per view.
• Output → a motion-ready interface where visuals communicate system reasoning.
V · Encode the Interface
• Translate design intelligence into reusable components.
• Name components after conceptual roles, not visual appearance.
• Bind each component to state rules defined in the model.
• Tokenize spacing, color, and timing so they can be coded exactly once.
• Pair with developers to validate constraints, breakpoints, and animation curves.
• Output → a working component library whose naming mirrors system logic.
VI · Test the Experience Mechanically
• Evaluate performance, coherence, and cognitive load.
• Run latency and feedback audits on real devices.
• Time every key interaction; identify drags beyond 300 ms.
• Check consistency between concept, copy, and behavior.
• Use heuristic testing focused on precision: clarity, latency, reversibility, and tone.
• Output → an audit log that treats UX quality as measurable mechanical feedback.

VII · Communicate for Implementation
• Make reasoning visible to collaborators.
• For every major decision, write a 3-line rationale: problem → principle → choice.
• Present interaction diagrams and timing tables in developer-ready form.
• Conduct critique sessions that target coherence and responsiveness, not aesthetics.
• Maintain a decision ledger that captures the evolution of the model.
• Output → shared understanding that keeps logic intact through production.
VIII · Maintain Conceptual Integrity
• Keep growth from breaking the model.
• Schedule monthly model audits — does the interface still match core concepts?
• Trace every new feature to a node on the original diagram.
• Refactor inconsistencies before they propagate.
• Archive motion and logic tests as regression references.
• Output → a continuously evolving but internally consistent interface system.
IX · Ship and Reflect
• Close the loop like an engineer.
• Compare the shipped build against conceptual and interaction specs.
• Record deltas: motion offsets, color drift, timing variance.
• Fix or document trade-offs and feed learnings back into tokens or models.
• Write a one-page post-flight report per release — what worked, what broke, what’s next.

Raghav Agrawal @impactology@mastodon.social@impactology· Oct 3

India – Affiliated College Model vs Western – University Model

Raghav Agrawal @impactology@mastodon.social@impactology· Sep 25

Most honest person I've met recently was a tech VP who told me: "I manage a team of twelve people who create documents for other teams who create documents for senior leadership who don't read documents" The hidden economy of nonsense by Alex McCann thestillwandering.substack.com/p/the-death-...