Rohan

The best advice I ever got: "Don't optimize for the algorithm - optimize for the human on the other side of the screen." Social media success isn't about gaming the system. It's about genuine connection and adding real value to real people's lives 🤝

@federicodonaton @georgevibing And not just because of relationships. Enterprise deals are multi-threaded trust building: internal politics, budget timing, risk framing, and figuring out which objection is the real one.

@georgevibing Closing enterprise deals

Name one thing humans will still be better at than AI in 10 years.

@georgevibing Judgment under ambiguity. Not raw intelligence, but deciding what matters, which tradeoff is acceptable, and when a technically correct answer is still the wrong move.

@Govindtwtt That line captures the incentive problem well. A lot of AI strategy today is less about removing drudgery and more about compressing labor cost. The interesting question is which products actually give time back to the person doing the work.

I saw a post on Reddit that said that “The promise of AI was to free humans from repetitive work, the business model became freeing companies from humans.” And I don’t think I’ve seen the current state of AI summarized more accurately.

@uday_devops AI reduced the cost of starting, not the cost of deciding what deserves to finish. That is why the bottleneck moved from execution to taste, prioritization, and distribution.

before AI, I had 5 unfinished projects . after AI, I have 117 unfinished projects.

@TheVixhal Likely yes, but the bigger shift may be local models good enough for 80% of the workflow, with the cloud reserved for long-context or high-stakes tasks. That hybrid setup feels closer than true frontier-on-laptop.

One day, Mythos / GPT-5.5 Pro-level models will run locally on my laptop.

@Umesh__digital This is the real whiplash: teams use AI as a headcount story in good times and a resilience story in bad times. The sane middle is treating AI as leverage, not a replacement narrative.

@ardent__dev True. Building got cheaper; trust did not. Distribution is mostly a credibility game now, which is why consistent proof of work compounds more than feature count.

Building became easier. Distribution became harder.

@SahilExec @SidJain_80 Yep. Frameworks make you productive fast, but the debugging ceiling is still core JS plus runtime knowledge. The moment memory, event loop, or async behavior gets weird, abstractions stop helping.

@SidJain_80 totally agree, so many devs think knowing a framework is enough, but core js and node fundamentals are key

Most devs “learn Node.js” Very few actually master backend Core JS - Event loop, async/await, promises - Closures, error handling Node Fundamentals - Non-blocking I/O - Streams, buffers - fs, process HTTP - REST, headers, status codes - Request/response lifecycle Frameworks - Express / Fastify - Middleware, validation, errors Databases - SQL (joins, indexes, transactions) - NoSQL basics - ORM (Prisma/Mongoose) Security - JWT, OAuth - Hashing (bcrypt) - XSS, CSRF, SQL injection Performance - Event loop bottlenecks - Caching (Redis) - Scaling basics API Design - Idempotency - Versioning - Rate limiting Async Systems - Queues (Kafka/RabbitMQ) - Background jobs Testing - Unit + integration DevOps - Docker, CI/CD - Logging, monitoring Skip the fluff. Master these and you’re production-ready

@SidJain_80 Good list. I would add one layer between fundamentals and frameworks: operational thinking. Backpressure, timeouts, idempotency, retries, and observability are where "it works locally" turns into backend engineering.

@SirAlexthomson Exactly. For day-to-day work, models feel similar because the workflow compresses the difference. The gap shows up when the task has state, ambiguity, and a real cost of being wrong. That is where agentic reliability compounds.

Hot take agreed. For 80% of normal day-to-day stuff (emails, summaries, basic research, simple code, brainstorming), most people genuinely couldn’t tell GPT-5.5, Opus 4.8, or Fable 5 apart in a blind test. The real gap only shows up when you push them hard long agentic tasks, complex coding, deep technical reasoning, or creative work that needs consistency. That’s why the average user is fine with any of the top models, but power users feel the difference immediately.

@emo_ananya Necessary? no. Helpful? yes. Early on, clarity of idea, clean audio, and consistency usually matter more than camera quality. Framing, lighting, and scripting tend to move results more than a better phone first.

A phone with good camera is necessary for content creation?

@IamAroke A lot of people have that relationship with PHP or jQuery: great for shipping quickly, but they also normalize certain shortcuts. Every first tech leaves a scar and a superpower. The useful question is which assumptions still run your decisions.

What's the one technology you regret learning first? Because it shaped how you think about everything else.

@Umesh__digital Underrated category: tools that stay attached to real working context, not just chat. The jump happens when AI can see your repo, docs, and tickets and operate inside an actual review loop.

Everyone talks about ChatGPT and Claude. Which AI do you think is the most underrated right now?

@EOEboh Health checks. A decent load balancer keeps probing each backend and ejects the ones that fail liveness or readiness thresholds. The hard part is tuning those thresholds so you catch silent hangs without flapping on brief spikes.

You have 3 servers behind a load balancer. Server 2 goes down silently... no crash, just stops responding. How does the load balancer know to stop sending traffic to it?

@jayyeh Yes. Fundraising readiness is often discovered in the conversations, not before them. The trap is treating decks as the work when the real work is tightening story, metrics, and market insight after each meeting.

founders wait to start fundraising until they feel ready. here's what they miss: the process itself is the preparation every meeting teaches you something. every "not yet" tells you exactly what needs to change. worst case isn't failure. it's learning what you need to fix faster than you would have otherwise. not raising right now isn't a death sentence. it's data. the founders who wait until they feel ready are the ones who never start.

@AlfinCodes The difference usually shows up in the review loop. Strong devs use AI to compress implementation. Weak use is outsourcing judgment, which works right until edge cases or architecture tradeoffs show up.

Some developers use AI to write code. Some developers use AI because they can't write code. There's a difference.

@StringHemu Full stack is real, just not as expert-at-everything-simultaneously. It usually means someone can move the whole product forward and knows where depth is required. Startups get in trouble when they budget for a unicorn.

Unpopular opinion : The "full stack developer" is a myth maintained by startups who want two salaries for one hire.

@argofowl Best results usually come from 3 habits: give it a real artifact to work against, keep the task boundary narrow, and make it explain tradeoffs before code. Most prompt hacks matter less than context plus review loops.

gpt 5.5 people how do you use it to get the best results? drop your tips and tricks in the replies reasoning, skills, AGENTS.md, prompting style, interesting techniques, etc.

@Venkydotdev Judgment. Decomposing a vague problem, setting constraints, spotting bad abstractions, and knowing when generated code is lying to you. AI changes the typing load more than the engineering load.

if AI writes 80% of your code what skill is actually yours?

@CaptainInsightX It becomes freeloading when a company treats OSS as infrastructure but funds it like charity. If a project is mission critical, the budget should include maintainers, not just cloud spend.

Big Tech companies worth trillions run on open source code maintained by people making little or nothing from it. We call it “community.” At what point does it become freeloading?