Michael Johnson

Dad, Co-Founder, Context Architect, Buchholz High School, Best Ideas Award winner (3rd grade, Norton Elementary)

[Michael Johnson @onemoremichael]

BYOH!

[Binfeng Xu @billxbf]

📜Paper: https://t.co/gYfCyIGLRJ
💻Codebase: https://t.co/vEXS9GfQKW

Thanks to the following contributors! @HaoZhang3438830 @ShaokunZhang1 @songyang_han Mingjie Liu, Jian Hu, @shizhediao Zhenghui Jin, @yunhengjackiez.

This work is done at @NVIDIAAI with great mentorship from @jankautz, @doyend, Michael Demoret.

The work draw inspiration from open-source contributions @slime_framework @sgl_project @radixark and @nvidia Megatron, RL and Gym family 💚.

[Garrison @HeyGarrison]

Do I know anyone at AWS who would like to collaborate on this? I love dynamodb.

[Michael Johnson @onemoremichael]

Dont fix it if it ain’t broken!

[Mitchell Hashimoto @mitchellh]

Fork your dependencies, trim them to only your use case, never update unless it breaks for your users. I’ve been vocal about this for 10+ years. I’ve always said that updating is way riskier than latent bugs (which can be tracked and CVEs monitored).

If you are updating a dependency, it’s on you to analyze every single commit in the full transitive set of dependencies. If you dont see anything compelling, dont update!

I remember at HashiCorp once in awhile an engineer would try to update a dep or replace a DIY lib with an external one and id always ask “show me the commit we need.” Dont update for the sake of it.

Feeling pretty swell about this mentality with all the supply chain attacks happening.

[Michael Johnson @onemoremichael]

@jasonfried But are the photos good?

[Michael Johnson @onemoremichael]

@jxnlco /goal implement in zig and then migrate to rust

[Michael Johnson @onemoremichael]

Shared state and pronoun resolution is such a unlock for voice enabled AI native UX

[Jarred Sumner @jarredsumner]

my favorite test failure during bun’s rust rewrite: TOML & YAML parsers stack overflow tests failed because it could now go deeper than the test expected

[Michael Johnson @onemoremichael]

@jarredsumner @popovicu94 @JordanHall_dev @wookash_podcast By far the most interesting subject on all this in my opinion!

[Michael Johnson @onemoremichael]

@allgarbled @championswimmer I’d say a similar anchor is the user and PM’s are well aligned for that as the comprehension North Star

[Michael Johnson @onemoremichael]

@sama @johnsimerlink I constantly get flagged doing this kind of work even tho I did the verification

[Michael Johnson @onemoremichael]

@tobi This got me feeling things… do I want to work at Shopify? 😂

[Michael Johnson @onemoremichael]

@badlogicgames Security work 5.5 gets flagged constantly, 5.3-codex is still functional

[Michael Johnson @onemoremichael]

@mattpocockuk And great docs/spec make the best foundation for shipping prototypes

[Michael Johnson @onemoremichael]

@buccocapital Right below your post on my feed…

https://t.co/TBBWex5Vaw

[Gokul Rajaram @gokulr]

POD-OF-ONE: THE NEW ORG BUILDING BLOCK

As a @coinbase board member, t’s been a privilege to watch @brian_armstrong @emiliemc, and the Coinbase team build a true AI-native company.

Brian's whole post is worth reading in depth. I want to focus in on one thing that Coinbase is testing: “one-person product teams.”

Most of the AI discourse has focused on one-person companies. The more powerful and more broadly applicable construct will likely be one-person teams inside companies.

The old product org split context across 3 people. The designer held the user experience. The PM held the customer and prioritization context. The engineer held the code and systems context. Coordination was the price you paid to combine those views into one shipping decision.

Agents reduce that coordination cost.

A single high-agency person can now ask agents to draft flows, write code, run QA, summarize customer feedback, generate variants, check edge cases, and produce release notes.

This model rewards a very specific kind of builder:

• Technical enough to inspect the work
• Product-minded enough to choose the right problem
• Tasteful enough to reject mediocre output
• Fast enough to ship before the org forms around the idea

The scarce skill is judgment.

One strong person with customer context and good taste can now do the work of a small pod. One weak person with agents just creates more output for someone else to review.

This changes how early-stage founders should hire.

The most useful hiring question is now: “Can this person own the outcome end-to-end?”

That’s a higher bar than a functional job description. It blends product sense, technical range, design taste, writing clarity, and operating discipline. The title matters less. The span matters more.

Call it pod-of-one thinking.

A pod-of-one builder can go from ambiguous customer pain to shipped v1 without waiting for specs, mocks, tickets, handoffs, or meetings. Agents fill in missing labor. The human carries the context.

Teams still matter. They should form when the surface area is real: multiple customer segments, production risk, complex GTM loops, or enough product depth that specialization pays for itself.

Before that, a pod-of-one may be the fastest shipping unit in the company.

Founders: hire people who can be pods-of-one, who can carry the whole problem in their head and use agents to increase their throughput.

[Michael Johnson @onemoremichael]

@mitsuhiko So many orgs laying off both the code sticklers slow to adopt AI tooling and the managers, when those are the actual skill sets I want with the new AI paradigm of work 🤦🏻‍♂️

[Michael Johnson @onemoremichael]

@snoopy_dot_jpg Same for driving school after I got a speeding ticket 😂

[Michael Johnson @onemoremichael]

@embirico This could apply here also

https://t.co/iPPZVIaEbn

[Michael Johnson @onemoremichael]

@thsottiaux Project-level memory profiles for Codex: each project can choose its own scoped memory context, or no memory, to prevent cross-project context creep.

[Michael Johnson @onemoremichael]

@thsottiaux Project-level memory profiles for Codex: each project can choose its own scoped memory context, or no memory, to prevent cross-project context creep.

[Michael Johnson @onemoremichael]

@TheEthanDing @tanayj Betting on a horse with only one or two winners but solid acquihire off ramps if your bet was a loser

[Michael Johnson @onemoremichael]

@jatingargiitk @Cloudflare Liability is going to create some very interesting case law. I wouldn’t be suprised to see some agent safe haven hosting countries start emerging