Product Engineer before it became cool, building https://xitter.gq/t.co/T1tbJTw8ZG, https://xitter.gq/t.co/4aOC3dOPKC, https://xitter.gq/t.co/CMzl30WGzj. I'm so much more than I tweet :)
object Object
Joined September 2018
- Tweets4.6K
- Following1.3K
- Followers1.3K
- Likes25.9K
Ejiro - retweeted
One of the hardest hitting Bible verses is, "Jesus Christ, the same yesterday, and today, and forever."
Imagine having to defend what Jesus said or did in 16 BC with, "it was the custom at that time."
2000 years later, no single spot on him by moral standards. Bruh!
Ejiro - retweeted
We're launching Bank Arena: 52 long-running banking tasks across several banking apps.
We gave frontier computer-use agents a bank to run, from recovering transfers to settling merchant payouts. The median task takes experts ~75 minutes to complete, and the highest-scoring models completed <31% of tasks.
Here's what we found ↓
Ejiro - retweeted
Introducing agent-desktop
agent-browser, for your whole computer
→ Computer Use for any coding agent
→ macOS, Windows and Linux
→ Runs in bg while you keep working
→ Agents use your computer in parallel, each with its own cursor
Ejiro - retweeted
danwo 0.1.6 is out.
• danwo now runs on its own proxy instead of a third-party one. A web install went from 257 packages to 54 packages
• The proxy speaks HTTP/2, so gRPC apps (Firestore, Pub/Sub, Temporal, your own services) work through it.
• The docs are now available in Simplified Chinese, starting with the quickstart and the fault matrix.
• If your config needs a package you haven't installed, danwo tells you once, with one command that installs everything.
• A page load that Chrome drops because the machine's network changed, which happens in CI when Docker containers start, is now loaded again instead of failing your test.
#AI #E2E #AIAutomation #testing
danwo.dev
Coding agents made code cheap. Checking it is the bottleneck.
Introducing dánwò: open-source e2e testing for the agentic era.
It drives your app like a user, then checks what happened behind it: one charge, one webhook, one email.
Then it breaks things on purpose to see what survives.
npx @ danwo/cli init
If you've ever shipped a payment bug your tests missed, I'd love to hear the story 👇
Fun fact: dánwò is Yorùbá for "try it and see."
#opensource #e2e #e2etesting #testing #devtools #typescript #qa
Ejiro - retweeted
It's very useful to understand the context around something, how others perceive it, what influences it, etc.
I have scraped together all evidence in chronological order, from prophecies before his time, to the surviving documents of his disciples enjoy
github.com/0xSero/all-histor…
Ejiro - retweeted
Getting @thallyio’s logo to show in Gmail sent me down a $990/year rabbit hole.
Got it working for $0. The missing piece was one Google profile setting.
Wrote up the steps for other solo and indie founders 👇 View article
Ejiro - retweeted
danwo 0.1.3 🎉
• Test Go & Python backends, not just Node
• Fault a single-page app's own API calls (500s, dropped connections, lost answers)
• Steadier on Linux, Windows fixes, security updates
End-to-end tests that judge what happens behind the UI.
danwo.dev/changelog
Coding agents made code cheap. Checking it is the bottleneck.
Introducing dánwò: open-source e2e testing for the agentic era.
It drives your app like a user, then checks what happened behind it: one charge, one webhook, one email.
Then it breaks things on purpose to see what survives.
npx @ danwo/cli init
If you've ever shipped a payment bug your tests missed, I'd love to hear the story 👇
Fun fact: dánwò is Yorùbá for "try it and see."
#opensource #e2e #e2etesting #testing #devtools #typescript #qa
Ejiro - retweeted
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Ejiro - retweeted
We also have them as test emulators in danwo.dev/docs/providers
Ejiro - retweeted
Coding agents made code cheap. Checking it is the bottleneck.
Introducing dánwò: open-source e2e testing for the agentic era.
It drives your app like a user, then checks what happened behind it: one charge, one webhook, one email.
Then it breaks things on purpose to see what survives.
npx @ danwo/cli init
If you've ever shipped a payment bug your tests missed, I'd love to hear the story 👇
Fun fact: dánwò is Yorùbá for "try it and see."
#opensource #e2e #e2etesting #testing #devtools #typescript #qa
Ejiro - retweeted
Introducing Eleven: reverse engineering with your agent.
i've always wanted to open any app and just ask what a function does, instead of staring at assembly for hours. so i built it
with Eleven you can:
→ open Mac apps, Android APKs, Windows and Linux binaries, and read every function as pseudocode, assembly, a graph or hex
→ ask Claude, Codex, Gemini, Cursor or any ACP agent about the code. it reads the binary with you and you watch every step
→ your agent renames functions and leaves notes as suggestions. you accept or skip, and everything can be undone
→ patch bytes safely (always a copy, never the original), compare two builds, and debug with lldb
free and open source
Ejiro - retweeted
We are excited to announce that Deno is joining Cloudflare!
deno.com/blog/cloudflare
Ejiro - retweeted
[L1nk in Thread]
Earlier this year, I launched a new creative portfolio with a developer. It was elaborate, heavily designed, animated, and I genuinely loved what we made.
Then I decided to build another one from scratch 😅
This time, I built it with Codex, Claude, and Antigravity.
The goal was simple: make something I could update on the fly, experiment with freely, and keep restrained enough that the work and case studies stay at the centre.
It is definitely not finished, and that is kind of the point. I’ll keep changing, breaking, and adding things over time.
I also brought one of my recent experiments into it, an orbital media gallery where projects float around an invisible sphere as a spatial archive.
If you’re on desktop, try Spatial Mode.
There truly is no better time to be a designer.
The distance between having an idea and being able to build, test, and put it into the world keeps getting smaller.
Ejiro - retweeted
Such an important point.
If things don’t compute, don’t make a judgement about your « smartness », you just lack primitives.
So, identify the primitives, focus on them and things will make sense.
#SysConf #SysConf26
Ejiro - retweeted
We're excited to announce Chowdeck (@chowdeck) as a Partner for SysConf 2026!
Chowdeck builds the technology and logistics behind on-demand delivery of food, groceries, pharmacy orders and packages across Nigeria and Ghana.
Chowdeck is a great example of engineering excellence. To run their app well, they have to solve on-demand delivery across a dozen cities: a hard systems problem that spans logistics optimization, ETA prediction, recommendations and fraud detection, amongst others. And they do.
Now, with their support, we're able to further our mission at SysConf to share knowledge that teaches, shows what's possible, inspires technical excellence and strengthens our engineering ecosystem.
See you at SysConf 2026!
#SysConf #SysConf26
ALT Chowdeck is a Partner for SysConf 2026.
Ejiro - retweeted
Say hello to Aster ✳
A coding agent you run yourself. Your key, your model, your machine. No telemetry.
withaster.dev
Ejiro - retweeted
Jev + Kimi K3 for fraud detection!
TLDR: Jev classified 100 emails in 1.42 seconds, then I routed the uncertain cases to Kimi K3. The full pipeline got 96/100 correct for only ~$0.07.
Video is not sped up, check out the live run!
Here was my process:
I gave Jev 100 emails to classify (a mix of 50 legit & 50 fraudelent emails). It classified all of them in 1.42 seconds.
An underrated feature about Jev is it will give you the confidence score for a classification, so I routed any prediction under 95% confidence to Kimi K3 to be fully sure.
31 emails fell below that threshold. After routing those to Kimi K3, the combined pipeline reached 96% accuracy.
The full run took 16 seconds & ~$0.07 in inference costs:
- $0.068 from Kimi K3 on @togethercompute
- $0.003 (1/3 of a cent) from Jev on @typesafeai.
I think this is a really interesting pattern: use a fast specialized model like Jev for the narrow task, then route the uncertain cases to a larger LLM.
I feel like this kind of approach could be a game changer for use cases like fraud or anything realtime. You can use the speed & low cost of Jev while having a larger LLM as a fallback to ensure high accuracy.