El Combo (@clcombe) profile picture

Going gray, worked in fin services for 20 years. Now in startup land, agentic engineering and product lead. Foodie, especially 🍙

Sydney
Joined March 2026
Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)Photo by El Combo (@clcombe)
El Combo retweeted
my AIE talk is up! come hear how the grok @bot team ships so quickly without compromise on quality youtu.be/O_zcc6Xl12g
370
777
213
8,726
1,483,119
Just pushed an update to my skills repo. All 19 skills have been improved with a total of 185 changes across them. Rerun the install command to update. aiforui.dev/skills
26
41
7
1,451
48,897
El Combo retweeted
pstack の /create-formal-verification-skill で作られる形式検証 Skill で形式検証のモデルと実装の整合性は担保しています 1. コードから対象を選び、/create-formal-verification-skill で形式検証 Skill を生成する 2. /statechart で図を出力し、人間が意図とのズレや漏れたり、壊れたりしているケースを確認する 3. 人間が指摘した漏れたり、壊れたりしているケースや意図とのズレを修正・実装させる 4. 生成された形式検証 Skill をエージェントに実行させ、性質が守られていることを確認する この Skill では、良い形式検証の原則を7つ定義しています pstack の考え方には「コードベースが究極の記憶」というものがありますが、この Skill もそれに忠実なので、振る舞いとモデルとの整合性は、pstack で担保しています 1. コードに忠実である(True to the code) 2. 失敗できる(Able to fail) 3. 見つけた問題は、問題として報告する(Findings stay findings) 4. 範囲について正直である(Honest about scope) 5. ツールを知らなくても読める(Readable without the tool) 6. 手間をかけずに実行し直せる(Cheap to rerun) 7. コードの変化に合わせ続ける(Kept in step with the code) また、この Skill ではコードから形式検証する対象を選ぶのと、コードを破壊して、形式検証 Skill を試すので、整合性が取れなくなることも確認するので、形式検証する場合は、現時点では pstack が最もベストな選択の1つとなります TLA+・TLC の結果を /statechart で図として読めるので、ここで漏れたケース・壊れたケース・人間の意図とのズレを指摘し、それらを修正・実装させ、最終的に形式検証Skillでエージェント自身に検証させています
Replying to @cu30rry_
丁寧にありがとうございます! 凄い、e2eが下に来るんですね… 形式検証は設計(モデル)の正しさを検証するものだと理解しているのですが、この設計とe2eが検証しているシナリオが一致していることはどうやって保証されてますか? TLAだとモデルと実装が別になるので、整合性の取り方に悩んでます
1
10
157
8,285
El Combo retweeted
if your product has a stupid on onboarding form on login please at least include a logout button so many products i login with the wrong google account and then i have to go clear cookies to escape the onboarding
33
4
1
606
23,027
Published a proper article on the Home Repo pattern - the simple solution to running agents across multiple repos. Contains a prompt you can run on your org's repos to help you set it up for yourself. Enjoy your new home aihero.dev/the-home-repo-pat…
84
101
16
1,389
71,275
El Combo retweeted
We just shipped multiplayer in Cursor Projects Everyone gets their own agents, working from one shared task board and shared context When your agent needs something from a teammate's work, it asks their agent directly
124
63
10
1,125
57,187
me using Fable 5.2 to center a div
69
390
39
5,794
282,701
El Combo retweeted
Your agents can now coordinate with your teammates’ agents. Projects has been rebuilt for multiplayer: every teammate and their agents can now work together in one unified workspace.
460
255
357
4,142
620,712
Narrative violation: Product managers are benefiting most in the AI era There’s been a lot of talk about AI replacing product managers (“who needs ’em!”). But so far, the data is showing the exact opposite. There are more open PM roles than at any point in nearly four years, up 75% from the March 2023 low, and PM openings have grown faster than any other core product role since the start of 2023. More here: lennysnewsletter.com/i/22420…
STATE OF THE TECH JOB MARKET IN 2026 Twice a year, I dig into trends in hiring and job openings. This is the fifth report I've done, and the most surprising yet: 1. PMs are winning in the AI era 2. Engineering demand has stalled 3. Design openings are at a record low 4. Every role now requires AI skills 5. Tech jobs are concentrating in the Bay Area 6. Remote roles are at a record low 7. The tech job market is growing fast, but mostly outside of software (gasp!) Full report here: lennysnewsletter.com/p/state…
66
32
21
411
37,650
El Combo retweeted
So bullish on MCP. Its utility exceeds the "M" part: even AI aside, it'll be a civilization-scale accomplishment to get all our systems communicating through one universal open protocol. A missing piece was how to get arbitrary callbacks across sessions. MCP could read and write, but it couldn't tell you later "this event happened." MCP Events fixes this. We're adding support in our own MCP. Say an agent makes a deployment, and wants to know when it's complete. Instead of polling, it can subscribe to 𝚍𝚎𝚙𝚕𝚘𝚢𝚖𝚎𝚗𝚝.𝚜𝚞𝚌𝚌𝚎𝚎𝚍𝚎𝚍. PSA: With Vercel Queues and Workflows, writing the backend machinery for your own MCPs to do this is easy peasy.
Vercel MCP now publishes events. Agents can subscribe to events like 𝚍𝚎𝚙𝚕𝚘𝚢𝚖𝚎𝚗𝚝.𝚜𝚞𝚌𝚌𝚎𝚎𝚍𝚎𝚍 and 𝚍𝚎𝚙𝚕𝚘𝚢𝚖𝚎𝚗𝚝.𝚎𝚛𝚛𝚘𝚛 and get webhook notifications when they fire. No more polling for changes. Changelog → vercel.com/changelog/vercel-…
168
64
20
1,121
89,327
Replying to @opencode
@opencode android app is out for everyone to use! Let me know what you think play.google.com/store/apps/d…
127
72
25
934
68,755
El Combo retweeted
Through internal testing, GitHub’s new Git storage architecture supported 5,000 concurrent agents working on a repo, sending 7,200 pushes per second, all while serving read traffic with consistently low latency.
119
119
75
1,149
106,851
El Combo retweeted
everyone relax android opencode app is almost done too reply below and i will dm you the testing link
827
32
5
1,736
50,534
El Combo retweeted
Introducing @𝚓𝚜𝚘𝚗-𝚛𝚎𝚗𝚍𝚎𝚛/𝚌𝚑𝚊𝚝 Generative UI for Slack, Teams and more → Agents reply with native cards → Tables, fields, charts and buttons → Streams in as the model writes → One catalog for every Chat SDK platform
130
199
60
4,622
278,949
El Combo retweeted
first 24hrs of boat CI: - 121 accounts, 229 repos - 19,172 jobs ran, 0 our infra couldn't provision - 7s median wait push -> running - 266 broken runners recovered automatically - 0 overloaded host due to CI users about 10% of our VM time spent last 24hrs was CI!
ciao blacksmith we just made a CI runner so cheap & fast you can scale your CI by 20x on the same budget introducing CI on boat.dev's cloud VMs 1. install boat & run: $ boat ci connect 2. change your runner: - runs-on: ubuntu-latest + runs-on: boat
18
3
2
134
8,389
El Combo retweeted
We find the users who send the majority of their messages in projects merge 6x more PRs vs non-projects users. Projects also helped teach me to delegate and validate the same way @bot does
Honestly if you're not using @cursor_ai Projects you're just missing a ton at this point. Don't sleep on it. Give it a try and see how it significantly improves your life. Here are some tips for you. I'm using it since it was debuted internally: * Have a single `Hub` that acts like a office manager. Use it for best practices, ask it to investigate how each project is working, what's missing? What can be done better? It there to help you and getter better * Start with a few projects first, don't overdo. With time you'll understand how many is enough. For example I recently started a debug session for some of our Cloud Agent nodes, I knew I was going to create multiple agents, review agents and so on. So I created a single Project just for this purpose * Make sure to use a proper model. Grok 4.7 is pretty good do the Project coordination part. I also use it for all kind of chores. If I need a big hammer, it dispatches agents with Opus/Fable. Cursor is nice because you're not bound to a single model! * Give more autonomy to your coordinators. Share that you "don't want to be bothered" , ask upfront if there are any MCP logins, or website auths needed for the work to be done. Remove all self inflicted blocking, and then let it rip. * You can assign nice icons and colors, choose them deliberately, it'll help during context switching. * Enable remote workers. By default Projects runs in Cloud Agents, but suppose there is something to be done that can be only on your localhost, the coordinator can dispatch an agent that runs in your machine. For more details read cursor.com/blog/projects and also my blog post arslan.io/2026/09/11/how-i-m…
1
1
131
El Combo retweeted
We're seeing significant improvements with our new inline common chunks feature! It's substantially reducing the number of small chunks in some of our internal projects. We're still refining it, but we've put together an in-depth guide explaining how it works: rolldown.rs/in-depth/inline-…
3
21
1
112
8,713