Dustin Deus
1.4K posts
Dustin Deus
@dustindeus
Building @opencode | prev: co-founded @wundergraphcom
Deutschland เข้าร่วม Ekim 2012
224 กำลังติดตาม853 ผู้ติดตาม
@blueemi99 @emanueledpt It's not so far away. However, instruct OpenCode to use the right tools to bench and evaluate in a loop. Bun has great profiling support.
English
In the next version of OpenCode, startup time improved by 38%.
github.com/anomalyco/open…
English
Code-as-search. Code Mode. All terms for the same thing. Let the LLM generate code on demand.
Put an agent in a tight loop with precise tools and shared vocabulary, and it outperforms any static system.
We’ll see more of this.
English
Totally agree. The next skill is understanding what’s worth building because with all this power, direction becomes the ultimate multiplier.
dax@thdxr
i have seen enough proof now that using a coding agent is a deep skill it's confusing because the people you see heavily using them produce horrible results but that's because it's a skill! you can get better and the ceiling seems pretty high - this is very exciting to me
English
I agree. I switched from VS Code and JetBrains just to give it a shot, and I’m still using it. 😁
arc@arcbjorn
Honestly feels like the pinnacle of code editor experience right now. It’s as light and blazingly fast on any moderate hardware as NeoVim when it’s needed. The Vim motions implementation is a chef’s kiss. So it totally captures old school senior engineers. It’s fully featured for the majority of mainstream languages as a true classic flagship editor when it’s needed — very easy to set up. Capturing the majority of engineers. And when needed, it rivals the best agentic coding editors with its own thread implementation and concurrent agents view. So now it captures non-technical folks — vibe coding looks pretty slick. Testing Zed since the private beta back in 2022 was an interesting experience. Seeing it grow is mesmerising. Best-in-class pair coding experience, best-in-class rendering performance, and the ability to just switch off AI when it’s not needed then switch back to a fully fledged editor. It truly feels like NeoVim 2.0. It’s virtually winning in most types of coding tool competitions (even competing with dedicated agent harnesses), except for JetBrains products which are also awesome and incredible feats of engineering, in a league of their own with a slightly different philosophy. It’s very inspiring to see the @zeddotdev team making the right decisions one after another, especially with the latest open source and license changes. Amazing job from super smart engineers.
English
I was looking for a speech-to-text tool and came across Hex, built by my colleague @kitlangton. It's very good and works locally github.com/kitlangton/Hex
English
Building a test harness with linting, benchmarks, telemetry, testing, and coding guidelines is one of the most underrated skills as an agent dev. Prove me wrong!
English
I pasted an api key into opencode and it told me to rotate it immediately.
Is this supposed to make it harder to read or something?
English
We self-hosted DeepSeek V4 Flash and pushed 600M tokens through it.
At a small scale, inference is about tokens/sec.
At real scale, it’s about KV cache, memory bandwidth, cache locality, and how much pain you can fit between L2 and L3.
English
@dustindeus @wundergraphcom was great to meet you yesterday!
English
I’ve joined OpenCode as a Principal Engineer 😎
My time at @wundergraphcom taught me a lot about developer tooling, infrastructure, product building, and what it takes to build for developers at scale.
Now it’s time to build again.
There hasn’t been a more exciting moment to build for developers in years. Developers don’t just need access to better models. They need strong defaults, reliable inference, tools that actually do the job, and the freedom to choose their models, own their stack, control their data, and optimize costs without getting trapped in someone else’s ecosystem.
I’m super excited to help make OpenCode the best way to build agents: from the best developer experience across the SDK and tooling, to great model choice and reliable inference for teams doing serious work with coding agents.
The pace is wild. The opportunity is massive. And this team’s ambition immediately pulled me in: @thdxr @jayair @fanjiewang.
English
I use Agents heavily, and when beginners ask me where to start, I usually give the same advice:
There are endless models, tools, workflows, and frameworks out there. It’s easy to feel overwhelmed.
Don’t.
The biggest mistake is trying to learn everything at once. Simplify first, then expand.
Here’s the approach I recommend:
1. Stick to one or two strong models.
For example: Opus or GPT-5.4.
Do not keep switching. In the beginning, consistency matters more than chasing every new release. You need repetition first, not constant novelty.
2. Build your workflow in a model-agnostic tool.
Use tools like OpenCode or Cursor so you can set up your workflow once and keep using it no matter how the model market changes. That gives you stability while everything around you keeps moving and improving.
3. Get good at thinking in requirements.
There is a reason many people say the next era of software engineers will be engineers with strong product manager instincts.
This is where the real leverage is.
The better you are at defining requirements, clarifying outcomes, and translating vague ideas into concrete plans, the more value you will get from AI.
Frameworks like skills or superpowers help because they improve context, sharpen deliverables, and turn fuzzy tasks into executable plans. A good framework is not just a prompt trick. It is a tool for clearer thinking.
4. Document every meaningful change.
AI generates code faster than humans can keep in their heads. Write down key decisions, architecture, and why they were made. Put the essential context into markdown docs or "AGENTS.md" so the next run does not start from zero. You can automate this too ;)
This helps future LLM runs, your teammates, and your future self. Without it, the same mistakes keep coming back.
5. Treat tests and verification as core parts of the workflow.
This is probably the most important point.
The way code is produced has changed. What matters now is less how code was written and more whether it meets requirements and creates value across correctness, architecture, maintainability, and performance.
Using LLMs does not mean giving up engineering excellence. In fact, this only works if we stay in control through deterministic signals like tests, static analysis, and validation.
That is also what makes agents effective.
These signals create the harness for an agent to operate in a tight loop, fix problems more autonomously, stay focused, and work within a defined scope instead of drifting. Without that harness, autonomy quickly becomes noise.
Once you understand this workflow, then experiment more. Try different models. Go deeper into context and prompt engineering.
But do not get lost in the details.
Learn the capabilities.
Learn the limitations.
Then build a workflow you can trust.
AI is not just about moving faster.
It is about building a system where speed, quality, and control work together.
English
Today, we are proud to announce the Open Source AI Manifesto: a grassroots set of principles for sustaining open source in the age of generative AI.
When AI scales code output faster than human attention can process it, that system breaks down.
We believe it is time to draw clearer boundaries around the role of AI in open source and to be explicit about the human responsibilities that cannot be delegated to a model.
🛡️ Humans must remain accountable for what they submit.
⚖️ Humans must verify, test, and stand behind their contributions.
⏱️ Humans must respect reviewer time.
🗣️ Humans must communicate with clarity and intent.
human-oss.dev
English
Dustin Deus รีทวีตแล้ว
Cosmo Connect is here! Federation for any API, not just GraphQL.
English
100%
Brian Armstrong@brian_armstrong
The first version of Coinbase launched with just a hot wallet - a risky proposition. We were in beta and the app prominently told people not to store any money there they couldn't afford to lose. But the amounts of deposits kept steadily rising. I realized we needed to build build a cold storage system to improve security (otherwise a single hot wallet breach would mean we were insolvent and the company would die), and called the two cryptography/security experts I knew (@zooko and @octal if memory serves) and asked them what the best architecture would be. They were super helpful and gave me a crash course, since I had never built such a system before. I asked them how long it would take to build and I remember one of them said it might take a team of ~10 people 18 months to get it all up and running and tested. The problem was we had about 8 weeks until the total deposits on the platform would exceed the total assets of the company, and only 2 engineers (including myself) to build it. We were seeing signs that hackers were already trying to break in, a true do or die moment. @satoshilite and I buckled down and set about coding the new cold storage system from scratch, and integrating it into the app. We made some reasonable trade offs but what we came up with was fundamentally secure, and a massive improvement. We even unboxed some new laptops for key generation, stored backup material across several safe deposit boxes and locations. With about a week remaining, we started the process of transferring funds over to the new system. We were both extremely sleep deprived (how mistakes happen!), and paired up to double check each others work as we sent over the first test transaction, then a bigger one, and so on until it was fully transferred. We breathed a sigh of relief and went home to sleep for about 12 hours. This was one of my proudest technical accomplishments from the early days of Coinbase: coding our v2 key storage system with 2 people in about 8 weeks, which should have taken 10 people 18 months. And it worked and served us well for years. We're now on ~v5 of key storage, and have advanced way beyond what we came up with that day. But if we hadn't gotten it out in time, Coinbase very well may not exist today. It's a great testament to how constraints breed creativity, top talent matters in startups, and teams are often capable of more than they think when there is no other option. Most products which succeed have early moments like this, where someone has to step up and make a play on the field that defies all the odds. As we face new challenges and deadlines across our many products, I always look out for who on the team is ready to step up and make the game winning play on the field.
QST
@thdxr That's just the beginning. It's currently a port, wait until they start integrating it with the language capabilities 😎
English
The more you integrate AI into your daily routine, the more you understand that its purpose is not to replace you but to give you superpowers. This is the era of Tony Starks 😎
GIF
English