BCGuy2010

"A market that /should/ go down, but doesn't, is actually quite bullish"

@StratLaboratory @jfsrev @threadreaderapp unroll

🧵 THREAD: ATR Extension - From Heuristic to Empirical P 1/ Backtested @jfsrev 8x ATR heuristic for the 50 SMA across ~2,700 tickers with nearly 5 years of data. The result? It's not just a rule of thumb. It's statistically correct. Here's what the data says 👇

@LyalinDotCom @geminicli I'm a subscriber to Google AI Pro but can't figure out how to access my $10/mo in CLI credits. And chatting with support isn't helpful.

Hey @geminicli fans, we'd love your help as we try to make it a great experience for you to give Antigravity CLI a try. Let us know how we can help smooth our the transition for Google One users.

@LyalinDotCom It's Jony. It'll be like the Apple Mouse. Under the car.

Right, looks cool, but until they 100% confirm how this thing attaches the charging cable no one is going to buy one.

@TedHZhang Thanks. How does GEV look to your expert eye? TYIA!

$EWY (Korean ETF) continues to chug along. After breaking out of this multi-month base, it is starting to flag up now. $KORU is the 3x ETF. This name looks very similar to $DRAM because Korea is essentially a memory ETF: Watch the group: $SNDK $MU $WDC $STX Sk Hynix and Samsung.

@CNBC CA #winning

@morganlinton Dude. You're trending again. "Developer Ditches Saturday Night Out for Pi AI Coding Tool" ... ¯\_(ツ)_/¯

And now kicking off my overnight build in Cursor w/Composer 2.5. So tonight I have three overnight builds running: - Pi w/GPT 5.5 medium effort working on a backpacking LLM - Codex with GPT 5.5 xhigh effort building an operating system - Cursor w/Composer 2.5 working on a new feature for PasteLocal And with that, it's almost 9:30pm, way past my bedtime.

@aleabitoreddit What’s your target on SIVE?

5 months ago. I gave a PT of $150 with $AXTI. Over 10x that current values from a $500M MC. It’s now a $9.2B MC and $140. $10 away. New $SIVE longs are starting to realize what’s it’s like to be in AXT. And that I’m pretty good at guessing intrinsic valuations of companies.

@ivanfioravanti Yes please!

I’ll have to create a version for Apple Local LLMs only 🤷🏻‍♂️

@morganlinton @stevibe @ChrisUniverse I’m interested in the keys!

Shipping these off to the lucky people who won them! Congrats again to @stevibe and @ChrisUniverse Still haven’t head back from the person who won the keyboard keys so might run another giveaway for those. As I’ve said many times, I’m not a big swag guy so excited to give these away to people who will truly appreciate them.

@getsmallai TY

@BCGuy2010 No definitely not, very different, Hermes does a LOT more than this. This is a very simple bare-bones harness.

New update to Small Harness, just pushed to Github. Built this feature with a combo of Cursor with Composer 2.5 and Codex w/GPT 5.5 Extra High 💪 The feature I added is a Ship Loop for local coding models, essentially the missing piece between “it edited a file” and “this is actually ready to commit.” Before: small 7B models had to guess. Run tests? Call shell. Check git? More shell. Multi-file refactor? Hope it writes JSON to disk correctly. Now in /mode ship they get first-class tools: - run_tests — structured pass/fail JSON, not a 256KB terminal dump - batch_edit — preview multi-file changes, then apply with approval - ship_status — branch drift, dirty files, conflicts, ready-to-ship in one call Built for ppl running small open-weight models locally who want an assistant that behaves like a junior dev on a real repo, not a chatbot with a bash addiction 🤪

@morganlinton If you had to guess, what’s your split between frontier token use vs local token use? For work projects and for personal projects? TYIA!

@BCGuy2010 Ty and yes, tested and really like Hermes, use it daily

I think I'm going to start sharing a list of what agentic coding tools me and my team are using, and what we're testing every week or so. As people have probably noticed, I'm a pretty energized and excited founder/cto. The reason why I'm so excited is that I think, in many ways, this is an engineering renaissance, and a time where engineers can benefit from doing truly interesting and creative work. I try to use any and every agentic coding tool that comes out, and ones that impress me, I go deeper with, and I tend to share my experiences on X. Once something really blows my mind, I typically start by rolling it out to my Senior Engineering Leadership Team, and if they love it, then the whole team. Today, we have a very clear and effective stack: 1. Codex + GPT 5.5 and Graphite for agentic code review: this combo is the current gold standard for agentic engineering today imo 2. Cursor: I'd say maybe half my team has been using Cursor, but with Composer 2.5, everyone now wants to use it. 3. New - Droid: I was so impressed with Droid in my testing, I am rolling it out to the team. We're going to do a one-month pilot where everyone can use it. I think this will be a huge unlock when it comes to optimizing for picking different models for different tasks, and running overnight missions. Now for my testing stack this week. This is stuff I'm evaluating, and a lot of what you'll probably see me tweeting about because I learned that when I test stuff, I guess other wacky ppl like me seem to like reading about it, so I keep sharing. 1. Grok Build: this has really impressed me and it's what I'm testing the most heavily now. It's changing daily, so that also makes it really fun to test because things that didn't work yesterday, work today, and new optimization I would have never thought of suddenly are available when I wake up in the morning. I'm like a kid in a candy store with Grok Build rn. I don't think it's at the point yet where I can move it to my Senior Eng Leadership Team to review, but I could see it getting there over the next few weeks. 2. Devin/Windsurf: I used this quite a bit last year and was comparing it to Cursor. I liked Cursor more in the end so switched. I know the team at Cognition has been working away making updates and improvements, and their new SWE-1.6 model is pretty neat, and crazy fast. I'm still experimenting with this one, too early to share much more than that. 3. Antigravity 2 + Gemini 3.5 Flash: this came out yesterday, so of course, I'm excited to try it. Tested out Antigravity when it first came out and didn't love it, but I'm always open to giving tools a second, third, or even fourth chance. Sometimes v1 just isn't a hit, and it takes time, I'm always ready to try something new. Gemini 3.5 Flash seems really fast, still to early to know much about the code quality but I will share as a I learn. I also want to test things like Pi, Zed, and a few others, but also can't really go deep testing too many things at once so I'd say those are on deck. Just know that when you see me super excited, talking about a coding tool I'm using, it's often something I'm testing vs. something tried and true that we already know and love. This means you might see me tweet a lot more about things like Grok Build, Devin/Windsurf, and Antigravity over the next week or two than something like GPT 5.5 in Codex which I already know are awesome. But I do think I probably could do a better job of sharing the tools that we already use and love...but I just get so excited about new things. What I'm playing around with could change weekly, it could change every two weeks, it will almost certainly change every 3-4 weeks. I am not dedicated to one model or harness, I just want the best for my team so they can do the truly fun creative engineering work that we all get to do in the Engineering Renaissance we are living in today. Phew, this got long, guess it should have been an article, but I dunno, just don't want to write that many articles, would rather just jam and put out a post like this, feels more natural. But I guess I'll make an image, with Grok, because that would be a fun thing to put in here. This is what it came up with, totally silly, but common, life is too short not to do silly stuff 😜

@KyleHessling1 @stevibe Is that temp range recommended for coding tasks? It’s not too high? It’s the opposite of conventional thinking? TYIA!

BREAKING! Qwopus 3.6 27B is LIVE! Thank you for your patience on this one, but I believe you'll find the wait was worth it! We've benchmarked this thing up and down, verified that it holds at least a 75.25% (152/202) in the initial 202 SWE bench solves. Not a full run of 500, but it shows the agentic coding quality from the original 27B is retained while adding all of the additional Qwopus benefits across many domains. As always, Jackrong is absolutely cooking here! COT quality has improved significantly through the inversion techniques from our Negentropy proof of concept. It also went through thorough curriculum training. You can check out the MMLU pro benchmarks on the model card, but it improved a whopping 10 points over the base model in physics, as well as meaningful jumps in Chemistry, business, and computer science. However, the best part is that I was able to build an entire survival shooter game using this local model entirely. I genuinely was blown away by the results, which you can play right now on my HF space (link in comments below). "Qwopus Commander" was completed in 9 turns of Qwopus 3.6! To test the new long context training, I made it re-output the entire 3000+ line program each turn, and it would make fixes and add features that I requested in large prompts, while perfectly replicating the entire rest of the game from context. What's more is that I did it all at Q8 KV cache quantization, and never had an issue over the entire 303k token run! IMPORTANT: Run it at --temp 0.75 to 1. Mess with it in that range for your use case. Higher temp actually lets the fine-tune shine and be exploratory and is also more stable. Swe Bench was run at temp 1, the game was built mostly at 0.8! We're so blessed to have all of you here and using the models! The support means so much! Please let me know what you build with it in the comments! Or if you have any issues getting it up and running, I will try my best to get back to you! Looking forward to seeing what you legends produce with it this weekend! huggingface.co/Jackrong/Qwopu…

@TedHZhang The speed of information keeps accelerating! I think this compresses even more so for liquid leaders. The stocks that folks like us want to buy.

$SNDK I use to always get uncomfortable buying short flags as they are often more failure prone. However, once you've identified one of the liquid market leaders of the cycle, they can work with far greater probability as institutions maintain a constant bid and support on them.

@XFreeze Not only is the new Gemini making amateur coding mistakes, the tokenomics are horrible. Three not hard questions into my new 5hr block and I’m timed out again.

SuperGrok Heavy is one of the best AI subscriptions you can get right now You’re not just paying for a chatbot You’re getting access to a growing xAI stack: • Grok 4.3 for text + reasoning • Grok Imagine for image/video creation • Grok Text-to-Speech + Speech-to-Text • X- search tools • Grok inside agents like Hermes Agent and OpenClaw • No separate API key needed for those agent workflows That best part is You can bring your Grok subscription directly into open-source agent workflows and start using it as infrastructure, not just an app

@sudoingX Are open weights out yet?

qwen is unreal. they just dropped 3.7 max and it is beating opus 4.6 max on most of the benchmarks they ran. terminal bench, mcp use, math, instruction following, humanity's last exam. and the apex math number, 44.5 against opus 34.5, that is not a small gap. the 35 hours straight on a kernel optimization task with 1000+ tool calls is the part i keep rereading. that is the agent era thing actually happening, not a slide. the speed alibaba is shipping at right now is the whole story, 3.6 was last month, 3.7 max today, nobody else is moving like this. one thing though, please open source this one too. 3.6 dense made the entire local llm ecosystem better. the max tier going api only would close a door we have been keeping open. give us the weights eventually.

@morganlinton Livermore? Interesting.

My favorite package to receive.

@morganlinton It sucks!

Now I'm getting confused. Does Gemini 3.5 Flash suck at coding, or is it amazing? 😵‍💫

@morganlinton Thanks!