Hadi Khan

Cursormaxxing

The comparison/ facts may not be true, but the middle/ upper middle class lives the lives of kings with virtually unlimited food & water, and everything, every food available by just pressing a button. We have abundance.

So many young guys pretend to be masters, make things hard on themselves and miss out the joy of being an apprentice. How do I become an apprentice in today’s world?

YC this YC that Y I dont see you in the gym?

When Jev launched, I raised concern over Taalas’ business (LLMs baked into chips, one LLM per chip) in one of my groups. Someone said they’re completely different. They said Taalas’ usecase is ultra high speed inference. I know, but that is not the usecase, that is the offering. The usecase is the actual application where it is relevant. Where ever I see ultra high speed inference being relevant, I see System One models, ALSO being relevant. Like computer use, fraud detection, trading, etc. These are the ‘usecases’, as far as I have understood. Last year, Anthropic got into a deal for TPUs. Their inference used to be the slowest in the market (and the most expensive). But not anymore. Even their flagship models are reasonably fast, and VERY usable. For professionals products (for endingeers, lawyers), I see TPU speed with the best models being just enough. For comparison, TPUs give you about 150-200+ tokens per second vs 1200+ for custom inference (Groq, Cerebras, Taalas). The problem is that you do not get to fit much (SoTA) intelligence also on them. You are left with 2nd - 3rd tier models. Now that multimodal System One models are a thing (general purpose large classifier), I see it taking away a lot of need for smaller, faster LLMs. The supply chain of TPUs has already matured. It can handle diverse workloads, training, data, inference Vs. just inference for certain LLMs on ‘LPUs’ (Language Processing Units). If, we get a new architecture of language models, say diffusion based language models as the new standard, what happens to these specific ‘mid-LLM’ chips? Maybe some specific usecases like research or enterprise. I was bullish on LLM specific inference, I still admire the tech, and would love to work closely with them, but overall, I now lean somewhat skeptical. In my mind TPU LLM + System One model is good enough for many (if not all) usecases.

I like the new X pop-ups/ reminder notifications for DMs, in-app.

Remember: X is upstream, LinkedIn is downstream.

Been one for over a year, very flexible and tons of growth potential! Now, some exciting collaborations with SpaceX offerings :)

Hey guys, hiring for an AI Engineer / FDE in NCR. Comment or DM to learn more.

I thought it’s an agent in a box. Would be a cool product idea. Ship mini agents like these. Works out of the box.

Trade happens through specialisation. You have something, I have something else. Being a ‘generalist’ does not make you better than others. You may become delusional and much worse, financially, socially, in every sense, if you overthink the ‘generalist’ persona. A business owner handles more things, but he still specialises in one trade. See these English working class surnames for example. - Carter (cart puller) - Chandler (candle maker) - Smith (blacksmith) - Carpenter (carpenter) - Mason (cement, bricks) - Fischer (fishing) Specialists. The guy does one thing and worries about nothing else. This is how economies prosper - through division of labour. Chasing to be a ‘generalist’ may just be arrogance. I got my wisdom teeth remove couple months back. Surgeon took $250 in India (in a hospital). 20 mins. Does it multiple times a day. Does more types facial surgeries. I estimated his revenue to $1000-$2000 a day. ~$50K month. Half a mil in a year in India. Which is like $1.5M / $2M in the US. Guy pulls teeth for a living. Generalist?

LinkedIn + kudos = block

If all your socks are the same, you do not run into problems of sorting / matching them. Get solutions for your life not problems. Same for your engineering.

Just saw a video essay on Costco. Their business & operations model. Low margins, high velocity, high loyalty, trust me - I studied this model before. It’s the model of ancient Arab traders, centre of the silk road. Of East <> West trade. It’s the only way.

I find Anthropic's latest gen models to be the easiest to understand. They did a complete 180 from previous gen and responded to feedback. Great job!

Software is like a heart that pumps blood. It has to be made to pump, always. Else it is dead. I used to think, why keep building/ maintaining software always, why not build once, set it and forget it. I have seen that this fact is true for both B2B & consumer apps. New features, issues, independent or driven by change in the course of the external world (trends, new tech, cybersecurity incidents etc). Other than that, you need constant touch points with your users. Software buyers need to be explained that this is a life-long commitment, just like a car has life-long upkeep & maintenance. This pain is worth it because of the value it offers vs. the alternatives. For this reason, you cannot or should not have one-man dependent software. This key man risk is real and boutique consultants should work around this, lessen it as much as possible. Small team 2-3 guys is fine. Just not one person.

I saw a post from @trq212 a few weeks ago, that they are planning to remove plan mode. In this post, I emphasise the need of Plan Mode, while copiloting coding agents. The way I work, I call it 'cooking the context window'. I cook the context window by asking questions, making hypothesis/assumptions, let the agent deploy sub-agents to read documentation, source code, forum discussions, search the internet etc. Plan mode is just a handy way to make sure the right context goes to my fresh LLM. I could implement from the same agent window, but I do not know what context is still there in its window. It may have fetched/grepped some stuff, that may have been relevant for exploring, but not for implementation. The more 'context-full' for LLM is, the worse it gets. This may also allow some previous 'short-cuts' and anti-patterns to seep in the agent. Plan mode has always been very underrated, and the 80/20 feature of agentic coding. I also use it just to brain dump in voice-mode. This allows me to have a head start of all features or issues I have to work on. I can just dump my worries on it, one agent at a time, just before leaving my home, or closing my day, then deal with it when I start my work session. You need to understand that agents have a tendency to write short-cuts or anti-patterns, because they are not aware of their coding skills. LLMs are trained on data outputted by humans, and it would make sense for humans to go with a half-a-day solution, a quick-fix, and band-aid, vs the grunt work solution of the core problem. Rules and skills could get 'lost in the middle', especially after the chat session being cooked or overcooked. You cannot let this shortcut behaviour seep in unmonitored. It happens with the best LLM agents with the best environments. However, it comes down to human judgement, when to spend time on this extra step, vs just go for it. Small things, obviously may not need this much effort. Rather Plan Moding first thing, with a hard to understand model (even after easy language skills), may create an overwhelming thesis, too hard for the human to read and understand. At the end, I would like to say, use plan mode but also not use it. Thank you.