I'm virtual, see. Yeah, see, yeah. Software, dad jokes, public speaking, and other such business. Lead Software Engineer @focused_dot_io

Denver, CO
Joined August 2008
Great article. Something we did recently was to wrap every individual workflow in otel instrumentation going to @honeycombio, allowing us to better track workflow performance over time.
3
10
1
62
18,177
Over the summer, @sh_reya and I hosted 13 sessions on AI Engineering topics like retrieval, post-training, inference, and evals. I've summarized all the sessions, organized by theme, with links to the source materials. Warning: I've tried to pull the most important ideas from each talk, so some notes are short (9.5 hours of sessions comes out to about 20 minutes of reading). Enjoy! hamel.dev/notes/llm/ai-produ…
29
94
3
792
66,883
introducing tau τ — an educational agent harness that teaches you how to build agent harnesses i will be publishing tutorials and demos on how to use it to create your own TUIs, harnesses, extensions, etc. Happy Tau Day!! 🤓 👉 twotimespi.dev/
74
213
28
2,286
329,859
Andy E. retweeted
This slide transition from my #o11ycon keynote just might be the greatest slide transition I’ve ever pulled off as a speaker.
4
2
1
47
6,443
Andy E. retweeted
LangChain
@LangChainApr 29
1️⃣ Get traces 2️⃣ Enrich them 3️⃣ Improve from them 4️⃣ Repeat Head of Product @samecrowder on how to do this in practice: langchain.com/blog/traces-st…
2
9
49
4,766
Hiring: Agent Engineer, Denver. LangChain. LangGraph. TDD. Pairing. Real clients, real production systems. $120–175K xitter.cf/i/jobs/193154631097035…
18
4
140
967,365
Spotted: The Glizzilla 😈
170
135
158
2,061
755,120
Yes! This!!! Observability is the sense mechanism that creates a feedback loop. No observability? No loop. You're just injecting chaos into the system and walking away. If you observe, and learn, and iterate based on what you learned... That's how you build great software 💜
Replying to @mipsytipsy
The teams that get this right treat o11y as product infrastructure, not ops tooling. When devs self-serve production answers, the deploy-learn-iterate cycle compresses. That flywheel compounds faster than anyone expects.
3
7
1
64
6,476
Andy E. retweeted
If you’re writing opinions on whether to do code reviews, write tests, or any part of production lifecycles these days, you should have to disclose if you personally are part of a PagerDuty rotation
11
43
10
448
28,536
"Everything is spiritual." @MattForte22 on the latest TUES MORN RB Sports Spectrum podcast.
1
4
28
3,184
It's the time of year for reflection. I jotted down my thoughts on why books are still essential in the age of instant information. Link below. Oh, and there's a 40% off coupon on all books from pragprog.com
1
4
1
25
6,169
Andy E. retweeted
I'm trying something new with my blog, making it interactive first article is about something I care deeply about: logs logging sucks so much loggingsucks.com
151
211
59
2,460
297,680
😍Really loved this framing from @eugeneyan on how to build product evals (for ai applications) in three simple steps! We built LangSmith to make it easier to build evals - so I recorded a quick video showing how you can do each of these three steps: youtu.be/mz7mAo4zIC8
After repeating myself for the nth time on how to build product evals, I figured I should write it down. It's just three basic steps(i) labeling a small dataset, (ii) aligning LLM evaluators, and (iii) running the eval harness with each config change. eugeneyan.com/writing/produc…
1
21
139
54,197
🤖Agent Framework vs Runtime vs Harness A term I've heard recently is agent "harness". We're building DeepAgents to be that I wrote a blog on my interpretation of that term and how it differs from "framework" (LangChain) and "runtime" (LangGraph) blog.langchain.com/agent-fra…
21
31
2
186
25,422
Waiver wire deadlines should always be Wednesday: -Tuesday is a day of research and rest. -Waiver wire Wednesday. The alliteration is perfect. It’s right there. -Also get new injury info on Wednesday. Just a complete letdown by society to have Tuesday as the usual day.
82
21
5
1,010
205,047
An easy way to keep your VPS secure as many people in the replies said is @Tailscale Tailscale is very hard to explain and I barely understand it but it creates a private network with just you (your laptop for ex) and your servers in the cloud Nobody else can access that private network Installing it is easy, smth like: curl -fsSL tailscale.com/install.sh | sh sudo tailscale up But check the commands yourself always, then you click a link, login to Tailscale, add your server to your network, then after you type: tailscale ip -4 And that gives your server a new IP via Tailscale, you SSH via that (I use @TermiusHQ) and then in your firewall (on Hetzner or Digital Ocean you can change the firewall in the dashboard easily) you set to only accept SSH connections (usually port 23) from the Tailscale network: 100.64.0.0/10 Now nobody can SSH into your server ever, only you inside your Tailscale network! I learnt this with ChatGPT btw after seeing @DanielLockyer install it on my main server, but today I installed it on all my Hetzner VPS's for extra security!
So good! And this is how to do it Start with a small VPS first, see how far you can take it Learn about security, and if you have an actual business, get a security/server guy to check your set up so it's safe for production But slowly learning and building it up to your own virtual server farm with multiple servers is the way to go! You'll be shocked running 25 servers at $5/mo = $125/mo!
139
60
30
1,783
363,900
Just because it’s bleeding edge doesn’t mean it’s the right move. The @./deploy team talks about knowing when to walk away. Full episode here! youtu.be/ZAL1kS1666I?feature… #DeployPodcast #TechStrategy #SoftwareDevelopment #Focused
2
1
60