Pinned Tweet
Announcing Hermes-node
Hermes can now run as a standalone CLI with Node.js-compatible APIs - no C++ embedding required.
Technically, hermes-node is a Node.js API compatibility layer for Hermes. It ports Node's native bindings (fs, net, http, child_process, etc.) to Node-API and lib/*.js files, allowing standard Node.js programs to run on Hermes instead of V8. It vendors libuv, c-ares, llhttp, Ada, simdutf, etc to provide the same networking and I/O stack as Node, and runs as a standalone CLI binary.
In this demo:
- REPL with tab completion
- TypeScript HTTP server using http, fs, and path
- Type annotations stripped natively by the parser — no transpiler, not even a built-in one.
This is still a work in progress, though surprisingly functional. More Node modules coming, source will be published soon.
Hermes can now run as a standalone CLI with Node.js-compatible APIs - no C++ embedding required.
Technically, hermes-node is a Node.js API compatibility layer for Hermes. It ports Node's native bindings (fs, net, http, child_process, etc.) to Node-API and lib/*.js files, allowing standard Node.js programs to run on Hermes instead of V8. It vendors libuv, c-ares, llhttp, Ada, simdutf, etc to provide the same networking and I/O stack as Node, and runs as a standalone CLI binary.
In this demo:
- REPL with tab completion
- TypeScript HTTP server using http, fs, and path
- Type annotations stripped natively by the parser — no transpiler, not even a built-in one.
This is still a work in progress, though surprisingly functional. More Node modules coming, source will be published soon.
29
55
503
78,620
Today anyone can generate a PR for Hermes with an agent. Sometimes people get disappointed if we don't accept it.
But think about this: if Hermes starts crashing, or becomes slow, or any problem appears, who is responsible for fixing it?
But think about this: if Hermes starts crashing, or becomes slow, or any problem appears, who is responsible for fixing it?
10
0
76
3,596
I can think of another reason why it is pointless to save money, but it is much more pessimistic 😅
@battleangelviv Saving money for the distant future is pointless with the advent of super intelligent abundance
3
0
10
1,612
Full Intl implementation for Hermes, based on ICU4X. Agents finally made it practical to develop and maintain internally.
It resolves all long-standing problems of the previous implementation, which was limited due to the inadequacies of the OS APIs.
github.com/tmikov/hermes/…
It resolves all long-standing problems of the previous implementation, which was limited due to the inadequacies of the OS APIs.
github.com/tmikov/hermes/…
8
6
102
4,480
I am working on approximately 17 things simultaneously, and frankly, I find it overwhelming, exhausting, and not very rewarding.
I used to find extreme joy in living with a problem. Keeping all the details in memory, thinking about it deeply for days, iterating and improving.
Now all of that is gone. I am still pretty good at guiding agents - decades of developing and shipping all kinds of software cannot be erased - but I feel that what I considered one of my biggest strengths is diminshing in value.
Anyone working seriously with agents knows this, even if they don't want to admit it. The deep super focused thinking is much more rare now. It is like a muscle - if you don't train it regularly, it gets weaker.
Is that something that we are willing to give up? Something that is no longer needed? Can the new generations of engineers do without it? No idea.
I used to find extreme joy in living with a problem. Keeping all the details in memory, thinking about it deeply for days, iterating and improving.
Now all of that is gone. I am still pretty good at guiding agents - decades of developing and shipping all kinds of software cannot be erased - but I feel that what I considered one of my biggest strengths is diminshing in value.
Anyone working seriously with agents knows this, even if they don't want to admit it. The deep super focused thinking is much more rare now. It is like a muscle - if you don't train it regularly, it gets weaker.
Is that something that we are willing to give up? Something that is no longer needed? Can the new generations of engineers do without it? No idea.
68
60
834
88,985
Agents are landing 100s of commits, but people realize that agents generate many more commits than needed, right?
Today I asked the agent to work on a feature. It produced 66 commits. These 66 should be 2 or 3 at the most. They have ZERO value as separate commits.
Also, agents always introduce many bugs. Those are discovered later, adding even more fix commits. Fixes, which in the "old ways" would have been merged into the original commits introducing them.
Agents increase productivity, but the number of commits is inflated, I strongly suspect.
Today I asked the agent to work on a feature. It produced 66 commits. These 66 should be 2 or 3 at the most. They have ZERO value as separate commits.
Also, agents always introduce many bugs. Those are discovered later, adding even more fix commits. Fixes, which in the "old ways" would have been merged into the original commits introducing them.
Agents increase productivity, but the number of commits is inflated, I strongly suspect.
19
0
51
5,313
A clarification: I wasn't asking for advice for how to deal with multiple commits. I am familiar with "squashing". I started using Git before many of you learned how to use a computer 😜
The point is that commit numbers are often inflated.
The point is that commit numbers are often inflated.
0
0
2
282
This was a fun interview at Agent Conf @reactuniverse_
"I only trust them when I don't look at the code." @tmikov is only half joking here. In the interview, he also explains how he reviews specs and tests, with Codex and Claude helping with review ⬇️
2
0
21
1,756
We need a new kind of PR - call it "problem solution". It only contains a description of the problem and a proposed solution. No code.
If your agent can generate the code, so can ours. This is actually easier than reviewing code, not to mention going back and forth with the submitter's agent. It is also far more secure.
The submitter should still get credit.
Only PRs where the code can't trivially be generated by an agent based on the description should contain code. Which I guess is almost none these days.
If your agent can generate the code, so can ours. This is actually easier than reviewing code, not to mention going back and forth with the submitter's agent. It is also far more secure.
The submitter should still get credit.
Only PRs where the code can't trivially be generated by an agent based on the description should contain code. Which I guess is almost none these days.
15
1
85
3,643
Tzvetan Mikov retweeted
Next year, I will assign one of my jira tickets to every kid in America.
This will be the first of many.
This will be the first of many.
104
455
11,089
217,514
Is this deliberate trolling for engagement, or are some people really this deluded? I can't believe anyone seriously thinks this.
2 years from now you will not work you and will not worry about money
you will get your cash payments from from all the major AI labs
think of it like 3-5 companies giving you a thousand dollar check every weekCommunity noteNo major AI labs have announced or committed to universal $1000 weekly cash payments or an end to work/money concerns by 2028. Discussions involve speculative proposals and limited small-scale pilots, not firm company plans.
telegraph.co.uk/business/2026/…
businessinsider.com/sam-altman-ubi…
bloodinthemachine.com/p/the-first-ba…
cdn.openai.com/pdf/561e7512-2…
2
1
23
1,539
I was talking to a good friend today about the future of our job - software engineering. He said something that made me pause:
The one thing that cannot be outsourced to AI is responsibility. Ultimately someone has to be responsible for features or bugs, and it can't be the AI.
The one thing that cannot be outsourced to AI is responsibility. Ultimately someone has to be responsible for features or bugs, and it can't be the AI.
136
30
731
48,197
My latest talk @nextappdevcon
Compiling a real-world CLI tool to native with Static Hermes speakerdeck.com/tmikov2023/com…
#speakerdeck
Compiling a real-world CLI tool to native with Static Hermes speakerdeck.com/tmikov2023/com…
#speakerdeck
3
1
40
1,595
Tzvetan Mikov retweeted
Achievement unlocked! Really happy to have spent the day with the legend @tmikov while at @nextappdevcon - thanks again for all your hard work and for helping make #nodeapi on #ReactNative more achievable than ever!
1
1
22
704
Tzvetan Mikov retweeted
I came across a Reddit post that described AI in a way I can’t stop thinking about:
“AI allows the wealthy to access skills while stripping skilled people of their ability to access wealth.”
I don’t think I’ve ever seen the purpose of AI captured so brutally and accurately.
“AI allows the wealthy to access skills while stripping skilled people of their ability to access wealth.”
I don’t think I’ve ever seen the purpose of AI captured so brutally and accurately.
366
3,678
16,155
298,108
Tzvetan Mikov retweeted
Yup, the problem with the fully automated luxury communism visions is that they miss power and what automating all labor will do to relative levels of it. Workers have power through their labor. If their labor is no longer needed, then they lose power and *may* get some handouts in return, but it's a fundamental and potentially irreversible change to the arrangement.
My book Obsolete came out last week (and is now widely available after excess demand cleared our first run). It takes both power and AI seriously. I'm so pleased it's finally out there! obsoletebook.org
My book Obsolete came out last week (and is now widely available after excess demand cleared our first run). It takes both power and AI seriously. I'm so pleased it's finally out there! obsoletebook.org
Humans having no more jobs isn't just about having no more meaning.
1. Your little job is the little bargaining chip with which people afford their lives. Being reliant on hand outs from the governments is a precarious situation, they can be revoked at any time.
2. People like Jon work in the media and inform people about what is going on, this supposedly also all gets automated. Jon has a little bit of power through his TV show. Automating all jobs => giving all power to AI.
3. Humans currently pose some threat to the government through violence or strike potential. Automating the military and security service will get rid of that.
This all is unlikely to matter as we will just look like dumb little bugs ("bootloaders" as e-acc calls it) occupying the precious real estate of this planet from the perspective of AI. But people are right to smell danger when they hear all labor will be replaced by these new creatures AI corps are building.
8
9
102
3,502