problem-solving

Bilingual

You don't have to be smart if you can think clearly
2 months ago
- Smart engineers rely on quick intuition but struggle when problems aren't instantly solvable.
- Strong engineers use methodical processes and can think clearly even at slow speeds.
- The key to clear thinking is focusing on invariants—true beliefs about the system.
- Thinking slowly and clearly is often unsatisfying but yields high value.
- Engineers who can think clearly under uncertainty are extremely effective.
Solution-space Taste
2 months ago
- The solution-space taste analogy uses a 2D field of obstacles to represent problem-solving in software design, where paths are solutions.
- Junior engineers often only focus on immediate obstacles, leading to overcomplicated solutions, similar to LLM coding agents that make drastic changes like editing React.js source code for minor errors.
- Experienced developers use local exploration and backtrack when they sense unexpected complexity, employing heuristics like a one-day implementation limit to signal insufficient understanding.
- Advanced practitioners scout parts of the solution space out of order to create stronger heuristics, connecting sketched pieces more efficiently based on prior domain experience.
- The analogy simplifies real-world complexity but serves as an intuition pump for navigating problem spaces and cultivating taste for clean versus overcomplicated solutions.
- Developing taste requires writing at least 100K lines of code, receiving feedback from better programmers, and carefully studying code written by experts.
What Languages Fix
2 months ago
- Kevin Kelleher proposed describing programming languages by the problem they fix.
- Many programming languages can be succinctly characterized in terms of the specific issues they address.
- Examples range from early languages like Algol and Fortran (fixing low-level assembly) to modern ones like Python and Ruby (fixing Perl's kludge).
- The approach reveals how each language evolves to solve perceived shortcomings of its predecessors.
Ideas for Startups
2 months ago
- Startup ideas are not million-dollar ideas; they are starting points that evolve through exploration and questioning.
- The best environment for generating ideas involves exposure to new technology and collaboration with like-minded friends, such as in a university.
- Ideas often arise from mental habits and doodling, requiring a real problem to focus the mind.
- To create a successful startup, make something people want by fixing broken products, commoditizing luxuries, or simplifying usage.
- Most startups are acquired, so consider building products that multiple large companies might want, and let ideas emerge naturally from fun projects with friends.
How To Choose A Subproblem
2 months ago
- Junior researchers struggle with long-term problems, needing to break 400-hour problems into smaller 1-hour ones.
- Subproblems can be created by asking simpler questions, studying simpler systems, or fact-finding.
- Toy problems and limits help simplify complex problems for easier analysis.
- Valuable subproblems provide actionable information, skill-building, and strategic insights.
- Graphs are high-information tools that often reveal surprises, even if not immediately actionable.
- Investing in human capital through learning is valuable but should be relevant to the main problem.
- Failing early isn't always optimal; persistence and patching can salvage promising approaches.
- Bad subproblems are vague, too hard (lacking a plan), too easy (yielding little insight), or useless.
- Easy checks can save time, even if they often confirm expectations.
- Seek external perspectives to avoid wasting time on overly hard or useless subproblems.
Three ways people respond to a problem (other than solving it)
2 months ago
- The consultant often gets hired after problems become severe enough that the cost of hiring is justified.
- Beyond solving problems, common responses include pushing problems around, preserving problems, and promoting new problems.
- Pushing problems around involves local optimization where solving one problem worsens another; blame should be on incentives, not individuals.
- Preserving problems occurs when institutions rely on problems for existence, so solving them threatens their purpose.
- Promoting new problems means that solving one problem can create another, requiring awareness of unintended consequences.
- Consultants must accept that problem-solving is ongoing and give up the illusion of finishing; balancing solving and ignoring problems is key.

problem-solving

You don't have to be smart if you can think clearly

Solution-space Taste

What Languages Fix

Ideas for Startups

How To Choose A Subproblem

Three ways people respond to a problem (other than solving it)