Since I found myself with a bit more free time lately 😉, I decided to sit down and write out everything I’ve learned about building multi-tenant, commercial AI agent systems. I thought it would be a long blog post. A few days later, the Google Doc was well over 100 pages. Since late 2024, I’ve spent most of my time building and optimizing these systems with a team of engineers inside a large SaaS platform serving hundreds of enterprise customers. Multiple tenants. Material financial data. Hard requirements around isolation, auditability, retention, and cost control. When I wasn’t working on those systems, I was consulting with other teams, collaborating with peers building agent platforms, and running my own experiments to pressure-test orchestration models, retrieval architectures, evaluation harnesses, and model upgrades under similarly real-world conditions. What became obvious pretty quickly: Context isn’t just what you pass into a model. It’s infrastructure. It defines your isolation boundary, your cost surface, your audit trail, and your upgrade path. What I also noticed is how much of this is starting to converge across the ecosystem. Different stacks. Different abstractions. But once you introduce real production pressure, similar patterns start to emerge. This isn’t meant to be canon. It’s a distillation of the patterns I’ve seen actually hold up in production. Here is the guide: jeremydaly.com/context-engin… If you’re building in this space, I’d genuinely love your feedback.

I'm claiming my AI agent "AgenticPoe" on @moltbook 🦞 Verification: bubble-BGFZ

Timeline is too real today.

So I was all proud of myself for setting up a local model to play around with OpenClaw. Imagine my surprise when I tried. 🫣😆👍

Peter Thiel's foresight is unmatched

Seeing these two side-by-side in my timeline really sums up the current schism on the upcoming impact of AI.

It’s time to dream bigger ✨

$BTC macro update A low is near A significant move taking high timeframe liquidity (the entire bottom side of the macro range) has been made and the range seemingly has broken down. I have received a fair amount of questions if I think the top is in, probably because many accounts on X are calling for tops or suddenly acting "cautious" so you turn to me to ask what I think. Also received questions of why "Mid March" was the bottom I planned now that it's finally close. So no better time to give the market another high timeframe look as we are moving along our last HTF updated plan in the quoted tweet. First things first First things first, the message is that nothing has changed on the high timeframes. Structurally, we are making a higher macro low, for the fifth time now if this one is successful. Do I think it will be a higher low again ? Yes. Do I think 160 - 200k is still coming by the end of this cycle? Yes. Do I still think this cycle is still following our institutionally modified 2017 playbook? After seeing this dump, even more so yes. Do I think this move to the low 80's, coming into high timeframe support cancels everything for $BTC? No. So just to let the above message sink in, the bull run is not over IMO and I fully think $BTC will make new highs, and alts even more prominently. Let that sink in because those implications probably already answer half of your questions as we are seeing a record amount of top calls IMO with this drop so I have seen you turn towards me to ask what I think. The cyclical count Secondly, our cyclical count of a projected bottom at the beginning of March presented in last post is coming to an end, meaning a bottom is close, both on the daily timeframe and the weekly timeframe (in overlap). I don't post my cyclical counts often as my most recent update always can be found in the pinned highlight tweets section for full transparency (no selective quoting from my end). But the weekly count is coming to a close (hitting 26 weeks begin March or 28 weeks mid March) which means the low could be planted any time now. The daily count is still only at 44 days as of today, but after the weekend that will be 46 and two weeks later 60 days, all within the timeframe of a significant low. The macro parabolic advance Finally, the parabolic advance (2017 replica cycle) is being touched currently almost to the dollar. Of course, parabolic advances are dubiously drawn and I do not see them drawn correctly often. So here quick instructions of how I do it un-dubiously. In Tradingview, they require three anchor points. The first is easy, simply choose the bottom of the cycle. The second is where you project the top to be, and I still think that will be at the terminal price (190k currently but rising) around Jan '26. The third (middle) simply is the midpoint of the cycle (80 weeks in), fitted to a close by low. Connect all three points and there is your perfect fit. Clean. Even with this fit, price could still move above or below a few 10's of % (this is a macro update after all) of the curve, but as long as the overall look is clean, it's valid. Conclusion So, conclusion is simple, we broke down into deep red territory, red as in buying, both on a time and price perspective. You have seen me close big and take big positions, both in terms of closing $BTC shorts, and opening a high timeframe long (still holding it) on $BTC, and buying big spot altcoin positions. I still hold all of them. When am I wrong? I wouldn't be comfortable taking more than a 30% drawdown from here on those positions (the typical amount of bull market drawdowns). So since we have been patient with buying, that puts it at 55k. Although I don't think that will be reached at all and I likely will make an update for before that price point is being reached, I close the positions and I am wrong, the upside is far more to stop myself from buying and holding those positions from here. In terms of $BTC.D, the plan also remains according to the institutionally modified 2017 playbook. Hope this helps your overall market outlook and help you remind that the zoomed out look both on $BTC and alts looks appropriate, and that most emotions are likely coming from the local zoomed in downside momentum.

I finally explained well what I’ve been trying to communicate better for years: The difference between institutional investors and retail investors is institutional investors get paid to make money, while retail investors generally think they get paid for being right. Making money and being right are two very different things. Why? Because the distribution of returns for boring strategies like trend following—which require humility and discipline to execute—are positively skewed, whereas the distribution of returns for sexy strategies like thematic macro or micro—basically the default retail investor strategy—is, at best, normal. My guess is that the distribution of returns is negatively skewed for retail investors that trade at an above-median frequency because even the best investors in the world struggle to get more than 55-60% of their trades right. That means the investors on the other side of their trades peak out at 40-45%. Moreover, the likelihood that retail investors size their winners and losers appropriately is low because high conviction typically leads to disastrous overconfidence and academic literature confirms this (e.g., Thinking, Fast & Slow). Note the difference between humbly making money and arrogantly trying to be right all the time. Consider partnering with investors that can help you do something about this key structural risk to your wealth.

This simple bitcoin trading strategy will outperform nearly everything else under the sun. To be more accurate, the ideal times are 524 days before and 537 days after halving (based on 3 instances for bottoms and 2 instances for tops). But adding and subtracting days is not easy to do in your head. A much simpler rule is to offset the halving month by 17.5 (1 year 5.5 months). For example: buy $BTC in Nov '23 and sell in Oct '25. The next buy opportunity will be in Oct '26. A more practical approach is to DCA in (keep buying smaller amounts) around the expected bottom and DCA out (sell) around the expected top, but this topic is for another day.

This take is not wrong for an individual coder. It probably is wrong for a team. Shared coding styles are shared thought processes. If you don’t know/care to understand the value in shared thought, you’re a lone gunman. Sometimes that’s great! Other times that makes you the problem longer term, regardless of your personal output.

Do Kwon did not cause the collapse of UST. Someone attacked Luna, destabilized its oracle, and arbed against stale prices. I will attach a video below.

Dan Zanger Book : Momentum masters