Bilingual

Show HN: Dockside – I turned unused space around the macOS Dock into a workspace
22 days ago
- Dockside is a customizable file shelf for Mac that works beside or independently of the Dock, offering drag zones, automations, and quick access to recent files.
- It supports a 14-day free trial and a lifetime license for $5.99, with features like offline privacy, no data collection, and Apple Notarization for security.
- Users praise it for boosting productivity, with reviews highlighting its flexibility, seamless integration into workflows, and responsive developer support.
The Mental Models I Use to Work with AI
22 days ago
- Invest in upfront alignment by making implicit assumptions explicit in the initial prompt to avoid unexpected AI outputs.
- Restart with better context rather than steering AI when initial outputs are far off; avoid compounding errors through path dependence.
- Equip AI agents with the same tools you use (CLI, MCP, browser control) to handle setup, testing, and instrumentation tasks.
- View early bad AI outputs as taste signals that help define creative direction, not just failures.
- Use visual references instead of lengthy prose prompts to guide visual outputs more effectively.
- Build a digital reference library to map your taste and provide raw material for AI to avoid generic results.
- Apply design principles to steer models away from slop and maintain unique, opinionated outputs.
- Use adversarial reviews with different LLMs to pressure-test outputs and reduce human bottleneck in review processes.
- Think ambitiously beyond MVP instincts as cheap execution allows for bigger, more innovative projects.
- Provide AI with tools to verify its own outputs (e.g., browsers, testing frameworks) to create internal feedback loops.
- Codify lessons from AI mistakes into memory systems (e.g., AGENTS.md) to prevent recurring errors.
- Recognize that LLMs reflect consensus, not judgment; your own taste and context are crucial for differentiation.
- Focus on clear direction and context over exact wording; systems matter more than prompt syntax.
- Be skeptical of impressive AI outputs in domains you lack expertise; verify to avoid misinformation.
- Understand the AI security lethal trifecta: combining private data, untrusted content, and external communication risks data exfiltration.
- Prioritize direction over execution as execution becomes commodified; mistakes in direction are more costly.
- Reframe SaaS AI tools as coworkers rather than just tools, explaining the shift towards chat-based interfaces.
- View loud AI debates as belief systems (like religion) rather than evidence-based arguments.
- See frustration with AI capability gaps as opportunities for new products and workflows.
- Evaluate SaaS ideas by whether a smart person with an LLM could replicate them easily; differentiate on brand, distribution, etc.
The Ugly Phase
21 days ago
- The Ugly Phase is a recurring low point in challenging projects, identified through personal experience with creating conference talks.
- It occurs when progress feels overwhelming, often near a deadline, causing doubt, frustration, and a desire to quit.
- Recognizing and accepting the Ugly Phase helps reduce meta-spiraling and emotional toil, allowing focus on moving forward.
- The phase is characterized by feelings of hatred toward the project, but it passes with persistence and refinement.
- Applying the concept to other hard, high-payoff projects, like building a house or unpacking, validates its universality.
- Strategies to navigate the Ugly Phase include recognizing it, cutting unnecessary scope, and refocusing on the immediate goal.
- Documenting progress with photos can provide perspective and motivation during the Ugly Phase.
- It is crucial to avoid dropping self-care activities like fitness during stress, as they support resilience.
- The Ugly Phase is often forgotten after completion, similar to forgetting the pain of childbirth.
- Celebrating success after emerging from the Ugly Phase reinforces achievement and prepares for future challenges.
You Feel Exhausted Without Having "Done" Anything Physically
21 days ago
- Decision fatigue explains why mental exhaustion occurs without physical activity, due to constant decision-making.
- The brain's willpower is like a muscle that can become depleted, leading to ego depletion where mental resources are exhausted.
- Studies, such as one on Israeli judges' parole decisions, show decision quality declines as more decisions are made throughout the day.
- Microdecisions accumulate and consume cognitive load, occupying working memory even when not actively thought about.
- The Zeigarnik effect indicates that unfinished tasks are remembered better, causing intrusive thoughts and impairing performance.
- Creating concrete plans for pending tasks can reduce mental strain by allowing the mind to stop monitoring unresolved decisions.
- Strategies to manage decision fatigue include reducing trivial decisions, externalizing tasks, and avoiding revisiting past decisions.
Artificial Adventures
24 days ago
- AI assistants are most valuable for reviewing code and finding bugs, with frontier models performing well.
- Refactoring tasks benefit from AI, reducing the cost of fixing design mistakes, but review can be challenging due to mixed-in incorrect changes.
- Writing code together with AI is problematic because AI makes poor decisions and often disregards instructions, though future harnesses may improve collaboration.
- AI is effective for one-off scripts and small tasks where output can be easily verified, but fails for complex projects like implementing board game rules.
- Search and verification tasks work well when precision is needed, but caution is required for unverifiable answers.
- Brainstorming with AI yields banal suggestions, and existing harnesses are limited, highlighting the need for better control and interaction methods.
- AI subscriptions offer good value, but token-based pricing may change cost-effectiveness, with cheaper models often misaligned or less useful.
- AI use requires adapting practices to manage mental models and verification, with potential for future improvements in model capabilities and tools.
Can I get some brutally honest feedback on a Chrome extension I built at 16?
19 days ago
- ReadBack saves your exact reading position on webpages, allowing you to resume where you left off without searching.
- It offers features like manual page saving, custom folders, favorites, searchable library, reading timeline, and detailed statistics.
- The extension works across most article-based websites, including documentation, tutorials, blogs, and research papers.
- It is lightweight, privacy-focused, and stores data locally in your browser, requiring no account or login.
- Reading statistics track metrics such as reading time, completed articles, goals, streaks, and session duration.
- The tool is designed for everyday long-form reading, supporting productivity with a clean dashboard.
- Developer information and privacy details are provided, emphasizing no data sharing with third parties.
Emacs and AI, the combination you've been missing
19 days ago
- The author describes frustration with switching between Emacs and AI tools like Claude for coding assistance.
- An epiphany occurs about integrating AI directly into Emacs using Claude to write custom plugins, despite the author's limited knowledge of Elisp.
- Claude successfully generates a functional Emacs plugin, demonstrating that AI can help customize Emacs to create an ideal development environment.
Show HN: Fast, native Mac file manager (filters, fuzzy find, 9 MB, no Electron)
19 days ago
- Filter files by type, date, and size simultaneously with preset categories like Images, PDFs, or custom filters.
- Browse two folders side by side in dual-pane mode, each with independent folders, filters, search, and history.
- Preview images or PDFs instantly by hovering or using Quick Look, and view thumbnails directly in the file list.
- Manage multiple tabs that remember their layouts; use keyboard shortcuts for tab navigation (e.g., T for new tab).
- Perform batch file renaming with options like Find & Replace or sequential numbering, and convert image formats (e.g., HEIC to JPG).
- Access a sidebar with bookmarked folders and mounted volumes, and use a command palette (A) for keyboard-driven actions.
- Integrate with command line via a shell function (e.g., wf()) to open folders directly from terminal.
How to Write an Email
17 days ago
- Start with the point: first sentence should state the ask, decision, risk, or update; avoid warming up or building suspense.
- Put bad news early: state problems clearly without hiding behind context or politeness.
- Use as few words as possible: delete filler, repetition, and softening language; every sentence must earn its place.
- Be literal: say exactly what you mean; avoid sarcasm, hints, irony, or coded language.
- Be specific: use names, numbers, dates, and facts instead of vague statements.
- One email should do one job: focus on a single purpose like asking, deciding, updating, escalating, or confirming.
- Make the next step obvious: specify who needs to do what by when, including deadlines for decisions or feedback.
- Separate facts, judgment, and recommendation: keep these distinct; clearly state uncertainty and avoid presenting guesses as facts.
- Write for forwarding: ensure the email makes sense if forwarded without extra context, with clear topic, names, dates, and decisions.
- AI is fine but avoid generic language: use AI for drafting or reviewing, but edit to sound specific and human; avoid clichéd phrases.
- Respect time: aim for emails that fit on one screen; if longer, put the punchline first with short sections underneath.
- Prioritize clarity over saving face: be direct and clear rather than vague or overly polite but useless.
- Keep threads clean: reply on the same thread only for the same topic; start new emails for new topics with clear subject lines and appropriate CCs.
- Use helpful subject lines: include action, decision, update, or risk to aid quick triage; avoid weak or vague subjects.
- Follow a default structure: subject with topic type, first sentence as punchline, facts, recommendation or next step, and owner/deadline.
- Use email appropriately: avoid it for long debates, unresolved issues, or high emotion; opt for calls or chats for speed, then send a summary email.
- Focus on reader understanding: a good email prioritizes fast comprehension over sounding smart, warm, or impressive.
I stopped watching my agents work and started listening
17 days ago
- Tri-State Relay Service (TSRS) was built to prevent multiple agents from interrupting with spoken updates by implementing a queue system where the user controls when updates are played.
- The initial version featured 'Focus' and 'Ready' modes, giving the user control over playback, similar to voicemail, ensuring updates don't interrupt unless allowed.
- A 'Live' mode was added for continuous playback of updates as they arrive, suitable when users are actively monitoring agents, reducing the need to manually trigger playback.
- To improve recognition, per-line voice IDs were introduced, allowing different projects (lines) to have distinct voices, helping users identify the source without parsing text.
- TSRS prioritizes user control with rules like mute and focus taking precedence, and ensures failed audio generation doesn't disrupt the system, making it reliable for continuous use.
Good Tools Are Invisible
16 days ago
- A good tool should be invisible, allowing users to focus on their work rather than the tool itself.
- Tool users often reframe a tool's shortcomings as enjoyable puzzles, confusing the feeling of cleverness with actual productivity.
- Tool choices can become tied to personal identity, leading to tribal signaling and blind defense of flaws.
- Productivity should be measured by actual output and efficiency, not by the engagement or cleverness felt while using a tool.
- Good tool design involves providing excellent defaults and minimizing unnecessary configurability, respecting users' time.
- A steep learning curve is a cost, not a feature, and should be justified by genuine productivity gains, not sunk-cost fallacies.
Fresh: A powerful terminal text editor and IDE
15 days ago
- Fresh is a powerful terminal text editor and IDE with instant startup and low memory footprint.
- It handles large files efficiently and supports remote editing over SSH.
- Features include split-panel review for Git with staged/unstaged/untracked files, side-by-side diff views, and export to Markdown.
- It can run multiple Language Server Protocol (LSP) servers per language with merged completions and per-project workspace detection.
- Supports named daemons, detach/reattach across terminal disconnects, and Hot Exit for persisting buffers.
- Plugins are compiled with OXC and run in a sandboxed QuickJS VM, with packages installed from an official registry.
- The status bar is configurable with elements like a clock and git branch, and themes are editable live.
- Includes a command palette with prefix routing, keybinding editor with conflict detection, and full mouse support.
- Available for installation via one-line curl script, Homebrew, winget, npm, or GitHub releases on Linux, macOS, Windows, and FreeBSD.
- User feedback highlights its speed, efficiency, multi-cursor experience, and intuitive design.
TK, or the secret to effortless writing (2024)
13 days ago
- Use 'TK' as a placeholder for missing words or ideas while writing to maintain momentum.
- TK serves as a built-in TODO checklist, easily searchable with two characters that rarely appear in words.
- TK is preferred over alternatives like 'TODO' or 'tc' to avoid false positives in searches.
The End of the Slack Tax
12 days ago
- Workplace chat tools like Slack act as a tax on attention to maintain company alignment, but AI may reduce this need by keeping work inherently aligned through a shared system.
- Slack's low cost per ping increased overall coordination volume, but AI can absorb work directly into a shared picture, reducing the need for constant pings and status updates.
- An effective AI strategy should shrink the alignment tax; if AI increases chat dependency, it may indicate poor integration and unnecessary coordination rather than productive work.
- As work shifts to AI-managed shared contexts, companies can become more focused, with communication centered on direction and judgment rather than tactical chatter.
The Definition of Done
12 days ago
- Software developers often misjudge when a task is truly done, leading to incomplete deliverables.
- A task is not done if it only works on the developer's machine, lacks testing or feedback, has an unmerged PR, isn't in production, depends on other work, or isn't accessible to users.
- The proper definition of done includes functioning on all supported devices, multiple testing and feedback rounds, merged PRs, completed team work, and live availability in production.
- AI cannot determine when something is done; developers must decide, and relying on AI for this risks job relevance.
- Done is defined by creating real-world value, not just by finishing work efforts.
Punch Yourself in the Face with Reality
12 days ago
- There are two ways to use AI: one leads to overbuilding without real user engagement, resulting in 'AI slop' and waste; the other leverages AI to speed up initial steps and focus on solving actual problems.
- Using AI as an excuse to build endlessly without user validation is a form of procrastination, magnifying pre-existing tendencies to avoid reality.
- AI hasn't made building a successful startup easier; it only accelerates certain parts, but core bottlenecks like risk-taking and resilience remain unchanged.
- The real challenges in startups and life—such as facing rejection, taking risks, and persisting without support—are still hard and unaffected by AI.
- The danger of AI is its potential as an escape, allowing individuals to feel productive while avoiding meaningful action and losing touch with reality.
- Success in the AI era depends on relentlessly pursuing truth by confronting reality, identifying one's purpose, and focusing on impactful, authentic work.
Be a Star or a Janitor
12 days ago
- Prioritize work that either no one else can do or no one else wants to do, being a star or a janitor.
- As a janitor, tackle miserable tasks to earn team gratitude and learn quickly, focusing on projects over ongoing maintenance.
- As a star, seek unique capabilities or comparative advantages to drive value, avoiding replacement-level or hum-drum tasks.
- Use this framework to prioritize at the margins, as small choices compound over time to shape role and reputation.
- Apply with discretion, avoiding misuse as an excuse not to follow manager instructions or devalue certain roles.
Be using a meta harness for agents
11 days ago
- Developers should move from using single agent sessions to meta harnesses for managing multiple parallel agents.
- Meta harnesses allow spinning up multiple agent sessions simultaneously, enabling work on multiple issues or codebases at once.
- Recommended meta harnesses include Conductor, Cursor, and Superset, each with unique features and flexibility.
- Meta harnesses are best for quick tasks like bug fixes, but may require traditional IDEs for larger features due to context switching.
- Adopting meta harnesses can transform software development workflows as agent capabilities improve.
OpenAI's first hardware product is the $230 Codex Micro macropad by Work Louder
10 days ago
- OpenAI launched its first hardware product, the $230 Codex Micro macropad, developed with Work Louder.
- The device is a limited edition, programmable mechanical macropad that features six 'agent keys' with colored lights to indicate agent status (idle, thinking, finished, error).
- It includes a dial, joystick, six regular keys for accepting/rejecting changes, and a push-to-talk button, all customizable via Work Louder's Input tool.
- The Codex Micro supports USB-C and Bluetooth, comes with low-profile switches (silent or clicky), and includes 32 additional keycaps.
- While designed for OpenAI's Codex agentic coding tool, it can be used with other applications through additional layers in the Input app.
You're Not in a Funk: You are not stuck. You are pointed the wrong direction
10 days ago
- The concept of feeling stuck is often a sign of being in the wrong direction or mode, not a 'funk'.
- There are two primary life modes: Expansion (engaging with the world, meeting people, collecting inputs) and Introspection (pulling inward, processing deeply, focusing on a few key ideas or people).
- Modes are not fixed identities like introversion or extroversion; they are situational positions that everyone alternates between.
- You cannot effectively run both modes simultaneously; they compete for the same mental and emotional resources.
- Life's productive stretches resemble terrain: expansion is the climb, introspection is the descent, with corners as transitions between them.
- The slopes (climb or descent) are straightforward and linear, while corners require skill and judgment to navigate.
- Missing a corner leads to a 'funk'—staleness from refusing to turn when the road changes direction.
- Each cycle has two corners: the top of the climb (transition to introspection) and the bottom of the descent (transition to expansion).
- Your weak corner is typically the one that turns you away from your perceived identity (e.g., an extrovert struggles to turn inward).
- Signs a corner is coming include: for the top of a climb, feeling drained after social interactions and collecting unprocessed inputs; for the bottom of a descent, restlessness, predictable thoughts, and comfort turning into laziness.
- Distraction-seeking is a common alarm at both corners, signaling it's time to check the road ahead.
- Use 'levers'—specific activities that helped you get unstuck in the past—to navigate corners intentionally, adjusting them to point inward or outward as needed.
- Levers often work both ways; small changes can redirect their effect (e.g., a solo trip vs. a business trip).
- Pulling a lever early at the corner maximizes its torque, while delaying requires a rescue operation with higher cost.
- With practice, you can shorten transition times from months to days, feeling the road bend before it happens.
- Avoid turning too early; the skill lies in sensing the turn and timing it correctly to maintain momentum.
- Navigating corners well compounds over time, leading to growth where each phase builds on the last, creating an upward spiral or 'helix'.
- Corners determine the momentum carried between phases; mastering them allows expansion and introspection to become more effective and aligned.
- A practical framework: know your current slope, watch for corner tells, maintain a lever list for both directions, and act early at corners.
- Start by reflecting on past instances of getting unstuck to identify personal levers, which can be applied in short timeframes.

productivity

Show HN: Dockside – I turned unused space around the macOS Dock into a workspace

The Mental Models I Use to Work with AI

The Ugly Phase

You Feel Exhausted Without Having "Done" Anything Physically

Artificial Adventures

Can I get some brutally honest feedback on a Chrome extension I built at 16?

Emacs and AI, the combination you've been missing

Show HN: Fast, native Mac file manager (filters, fuzzy find, 9 MB, no Electron)

How to Write an Email

I stopped watching my agents work and started listening

Good Tools Are Invisible

Fresh: A powerful terminal text editor and IDE

TK, or the secret to effortless writing (2024)

The End of the Slack Tax

The Definition of Done

Punch Yourself in the Face with Reality

Be a Star or a Janitor

Be using a meta harness for agents

OpenAI's first hardware product is the $230 Codex Micro macropad by Work Louder

You're Not in a Funk: You are not stuck. You are pointed the wrong direction