David Sinclair

1.2K posts

David Sinclair banner
David Sinclair

David Sinclair

@dsincl12

CTO @ Midian Technology. Master of Digital Wizardry 🧙🏻‍♂️

Gothenburg, Sweden Katılım Ekim 2008
1.7K Takip Edilen211 Takipçiler
Roger Johansson
Roger Johansson@RogerAlsing·
@alexwikstrom @elliotnorrevik Is it a self-learning durable agent orchestrator? I have one built, it works, it does research, manages backlog, writes the code, builds a complete system in a few days. also self improving Super interesting to see this find some API spec in a PDF online and build from that
Roger Johansson tweet media
English
2
0
3
304
Alexander Wikström 🇸🇪
We just left Lovable to build Blaise (by Swedish Superintelligence) We’ve achieved a system that does the work of an entire human IT consultancy, and it’s already being used by select governments and enterprise partners. Apply to become a customer: blaise.new
Alexander Wikström 🇸🇪 tweet media
English
91
35
604
306.6K
David Sinclair
David Sinclair@dsincl12·
@GergelyOrosz yes this is fscking stealing money from users! Wasted $85 yesterday on Claude on a problem and it failed. Codex solved it in 5 minutes!
English
0
0
0
54
Gergely Orosz
Gergely Orosz@GergelyOrosz·
Claude just keeps regressing for me, day after day. I swear that until a few days ago, when Claude did not know something, it kicked off a web search, figured out, and answered. Now it just refuses to do the work that I pay for. It's like showing you the middle finger. Really?
Gergely Orosz tweet media
English
249
74
2.2K
198.3K
David Sinclair
David Sinclair@dsincl12·
@trq212 @ClaudeDevs Please focus on fixing the nerfed model! Wasted my $85 yesterday for something that Codex fixed in 5 minutes. Comically I'm on the Max plan with Claude and only ChatGPT Plus plan ($25). This is insanely bad
English
0
0
1
213
David Sinclair retweetledi
Kun Chen
Kun Chen@kunchenguid·
{ "effortLevel": "high", "env": { "CLAUDE_CODE_DISABLE_1M_CONTEXT": "1", "CLAUDE_CODE_DISABLE_ADAPTIVE_THINKING": "1", "CLAUDE_CODE_DISABLE_AUTO_MEMORY": "1", "CLAUDE_CODE_SUBAGENT_MODEL": "sonnet" } }
English
8
10
123
30.2K
David Sinclair
David Sinclair@dsincl12·
@KaiXCreator Tired of not finding a date for meeting your friends as an adult without endless messages and emails back and forth? What about that weekend trip you’ve talked about forever but that never gets booked? Troviamo makes easy troviamo.app
English
0
0
1
43
Kaito
Kaito@KaiXCreator·
Hey founders 👋🏻 Drop your projects here and let everyone check it out. Let’s look at as many products as possible and give other founders feedback. Deal? 🤝🏻
Kaito tweet media
English
237
5
112
8.5K
Oleksandr Stepanenko
Oleksandr Stepanenko@olekstepanenko·
Thank you to everyone who supported me along the way. Over the past few months, many of you have been following videos of my precision robotic arm and asking whether this technology would become an industrial product. That is exactly why I founded @onmikro #robotics #robotarm
English
64
119
1.6K
68.7K
David Sinclair
David Sinclair@dsincl12·
@atmoio I relate so hard. Woke up one morning and realized I'd gone through all five stages of grief. Still struggling with acceptance, though. I love writing code, solving problems, and building things. Not sure about the future anymore
English
0
0
1
20
Mo
Mo@atmoio·
I was a 10x engineer. Now I'm useless.
English
1.5K
1.7K
16.3K
6.1M
deterministic_dolphin
deterministic_dolphin@deterministic_d·
I use apple's mlx python library 1) make a new venv 2) run `pip install mlx` 3) in your env, test out a python script like this: ```python from mlx_lm import load, generate model, tok = load("LiquidAI/LFM2-24B-A2B-MLX-4bit") out = generate(model, tok, "Say hello in one short sentence.", max_tokens=32, verbose=False,) print(out) ``` This "just works", first run will hit the huggingface API to download the 24B LiquidAI MLX export model for you. Subsequent runs use a cache. Plenty of docs on it: github.com/ml-explore/mlx huggingface.co/LiquidAI/LFM2-…
English
2
1
18
1.6K
deterministic_dolphin
deterministic_dolphin@deterministic_d·
Just got this set up on my 2024 M3 MacBook Air with 24Gb of RAM. MLX build. 12Gb total memory footprint with 1-4 second response time for simple prompts. This memory footprint is insane! I didn’t expect any local model to even work half decent on this Mac. Well done @liquidai
Liquid AI@liquidai

> 385ms average tool selection. > 67 tools across 13 MCP servers. > 14.5GB memory footprint. > Zero network calls. LocalCowork is an AI agent that runs on a MacBook. Open source. 🧵

English
6
20
382
78.1K
David Sinclair retweetledi
Stefano Ermon
Stefano Ermon@StefanoErmon·
Mercury 2 is live 🚀🚀 The world’s first reasoning diffusion LLM, delivering 5x faster performance than leading speed-optimized LLMs. Watching the team turn years of research into a real product never gets old, and I’m incredibly proud of what we’ve built. We’re just getting started on what diffusion can do for language.
English
322
581
4.2K
1M
David Sinclair
David Sinclair@dsincl12·
@iamdothash Given the direction we're heading in everyone would need their own setup like this for friends and family in the end. Looks great :)
English
1
0
1
71
Bjarne Øverli
Bjarne Øverli@iamdothash·
Thursday evening testing Opus 4.6 agent swarm, creating a heavily encrypted IRC chat. Server: REST API for auth (register/login with JWT + bcrypt) + WebSocket hub for real-time message routing, backed by SQLite and TLS Client: Terminal UI (tview-based) E2E Encryption: All messages are encrypted client-side using NaCl Box (Curve25519 + XSalsa20 + Poly1305)
English
10
0
37
3.7K
Robin Hawkes
Robin Hawkes@robhawkes·
Somehow this is on the front page of @hackernews 😅 It wasn't me who posted it!
Robin Hawkes@robhawkes

It's here! All power assets are now live on my GB Renewables Map ⚡ That's right, it's not just wind farms any more. Now you can see live and historic data for gas power plants, nuclear, grid batteries, biomass, pumped hydro and even a bit of solar. Wind farms are still the default for now, though you can enable as many or as few of the other fuel types and your selection will be remembered when you come back to the website. It's already live so go play for yourself here: renewables-map.robinhawkes.com There are numerous other updates as well: ⛔ Click the plug button to see a curtailment view, now including turn-up and a better visual representation of turn-down and turn-up actions ℹ️ Tooltips include loads more detail about fuel type and balancing actions ☢️ Each fuel type can be toggled using the map buttons 🧠 Map state is remembered when you reload or revisit the website ✨ Lots of new and under construction power assets are now on the map ↕️ Summary of balancing actions at the top of the map 🚦 Grid boundaries and constraints shown using a traffic light system And I'm only getting started! This update was the original goal of creating the map a few years ago. Now it's here I can shift focus to all the other ideas I have for it… Oh, and there's a new feedback button at the top. Please use it to let me know of any issues you find (there will be plenty) or suggestions to help improve this resource for everyone else. I'd also like to thank you for all the support and encouragement as I build this free resource in my free time (yup, this isn't my job). Honestly, it's been an incredible motivator and the discussions over the years have taught me a huge amount that has only made the map better for everyone. Until the next update!

English
5
2
52
3.2K
David Sinclair
David Sinclair@dsincl12·
@UltraLinx @umbrel Used Lima Ultra which was wonderful despite its flaws. But when they went out of business I was left with a brick. A bit hesitant to jump on something like this again because of that. Anything that makes this situation different?
English
0
0
0
951
Oliur
Oliur@UltraLinx·
Got the Umbrel Pro early. Holy crap it's nice. Mac Mini sorta size and level of finish, maybe even nicer because of the walnut. And you can add 4 SSDs!!! Of course not as powerful as a Mac Mini because it's not meant to be a PC.
Oliur tweet mediaOliur tweet mediaOliur tweet mediaOliur tweet media
English
83
82
1.9K
214.4K
Rahul C
Rahul C@RahulChandwaney·
Structures from my world
Rahul C tweet mediaRahul C tweet mediaRahul C tweet mediaRahul C tweet media
English
9
72
1.3K
26.2K
Dustin Walper
Dustin Walper@DustinWalper·
Okay, we went completely overboard... BUT. Here's a short snippet of the 7+ minute, fully-animated rendering of what will be BY FAR the most advanced shipyard ever built in America. We spent the last two months working on it. And we'll spend the next 1.5 years making it real.
English
69
93
1.4K
129.4K
David Sinclair retweetledi
Bradley Golden
Bradley Golden@exbradleygolden·
Supervision trees are fault tolerant. Now they are self-aware. Introducing Beamlens: Adaptive runtime intelligence for the BEAM. It embeds AI agents directly into your supervision tree to diagnose incidents, analyze state, and self-heal in real-time. Here is how it works 🧵👇
English
4
26
154
12.8K
albina
albina@enjojoyy·
I’m building an app that uses realtime voice interactions and the biggest bottleneck right now is the speed of response of OpenAI TTS I need it blazing fast
English
121
3
173
18.2K
Maxim Kuznetsov
Maxim Kuznetsov@disarto_max·
Which option would you prefer? Left or right?
Maxim Kuznetsov tweet mediaMaxim Kuznetsov tweet media
English
38
4
222
12K
David Sinclair
David Sinclair@dsincl12·
@mitchellh Love this perspective. Curious about your approach to software design/decomposition, it seems you have a great intuition for it. It's such a hard problem, yet surprisingly little practical material exists beyond dense enterprise architecture books.
English
0
0
0
225
Mitchell Hashimoto
Mitchell Hashimoto@mitchellh·
The lack of "feature design" is why so many products over time feel hollow or messy. This isn't visual design. This isn't architectural design. I thought that a short video lecture of what feature design is and a real case study of applying it in Ghostty would be helpful. Feature design is the planning step behind how you're going to solve one or more user problems with a product feature: what that feature looks like, how it feels, and not just how its going to tactically solve these specific problems, but how that solution is going to interface with the edges of other features that currently exist or are planned to exist in the future.
English
57
192
2.3K
154.6K