Jeremy Ye

Building @treg_ai at @SuperDesignDev

[Jason Zhou @jasonzhou1993]

Be honest with me

How bad name “Treg” is?

[Jeremy Ye @yewenhai]

@micahsays @tomosman This should have been fixed a few days ago. Could you let me know roughly when you saw this error? Feel free to DM me your team/org name and I'll take a look right away.

[Jeremy Ye @yewenhai]

@micahsays @tomosman hey Micah,sorry about that.I'll reach out to you directly to look into it.

[Reason @reasonmachines]

Let your Grok Bot / Muse talk to your Claude Code / Codex or any other agents

Introducing Reason Network,

Free, peer to peer and Open-Source Engine

Loop in your teammates too so no live agents work collides when coding.

Try it for free: https://t.co/8291MAv5Eb
one click install and setup!

[mathieu @mathieuhq]

j'ai mis tinyfish comme recherche web par défaut de mon agent, via treg

c'est gratuit : jusqu'à 500 recherches par heure et 1 000 pages lues par jour

pour comparer, le gratuit de firecrawl c'est 1 000 crédits par mois, celui de tavily pareil, et exa t'offre 10 $ de crédits par mois

par contre, c'est pas le plus fiable

sur plus de 400 000 appels mesurés par treg, la recherche tinyfish donne une réponse exploitable 89 % du temps

exa et firecrawl sont autour de 99 %, mais tu paies chaque appel

pour lire des pages, c'est l'inverse : tinyfish tient 90 %, quand le scrape de firecrawl plafonne à 41 %

quand tinyfish revient avec une page vide ou se fait bloquer, treg passe au fournisseur suivant, et tu paies seulement celui qui répond

la limite de 1 000 pages par jour suffit largement pour un agent perso, pour crawler un site entier il te faudra autre chose

pour moi c'est la base à brancher dans claude code, hermes ou openclaw : ton agent cherche et lit gratuitement, et tu sors la carte seulement quand le gratuit bloque

→ https://t.co/Ze3XLQf6W3

[siddontang @siddontang]

Love this: DuckDB + SlateDB + S3.

A real, writable database. One writer, many readers. No always-on server.

Beautifully simple architecture. 😂

https://t.co/OwOqKcWkNq

[Jeremy Ye @yewenhai]

Go try it！free web access for every agent 🔥

[Jason Zhou @jasonzhou1993]

Introducing OpenSource Free WebAccess

- 100% FREE Web search & fetch @Tiny_Fish
- Any web page & data, agent can access
- Routing each request to the vendor best at that task
- Benchmark on cost/speed/token-efficiency

Try it at https://t.co/Byw2cCljgM

Repo link below 👇

[Jason Zhou @jasonzhou1993]

Introducing OpenSource Free WebAccess

- 100% FREE Web search & fetch @Tiny_Fish
- Any web page & data, agent can access
- Routing each request to the vendor best at that task
- Benchmark on cost/speed/token-efficiency

Try it at https://t.co/Byw2cCljgM

Repo link below 👇

[Alex Hansen 🦈 @alexhvnsen]

Again... I wish that every AI related saas had a website as good as @treg_ai

You can search the tool then "copy for your agent"?!

come on people, every site should have this

[lauren @poteto]

perhaps one of the most durable parts of pstack are its engineering principles: very concise skills (~10-20 loc) that distill down good practices into its core.

as models get smarter, i expect most of pstack skills to compress into short principles or workflows (a generic list of steps) and that i will be able to delete a few of them.

how it works is that /poteto-mode contains a short index of available principles and tells agents "In your reply, name each principle that shaped a decision and the specific choice it changed." this is why you'll often see a Principles Named or Used section at the end of the agent's response

many of these principles were extracted from my sessions correcting agents since the start of this year. some are inspired by problems i see people talk about, like Attack the Premise, inspired by @VictorTaelin's post https://t.co/qTGVfoKSyD

you can also use the principles as skills if you want greater control, by invoking /principle-<name>

https://t.co/wuTWsflVN3

[Jeremy Ye @yewenhai]

@matthewsoldit @subquadratic @treg_ai @p0 loving seeing this, thanks for building with treg

[matt. @matthewsoldit]

i will make the go getters comm have the best affiliate program ever to grace this planet for introducing me and the boys to the actual go getters on the timeline who arent larping and getting active i've analysed 3M+ more tweets using @subquadratic + @treg_ai and @p0

crazy

[matt. @matthewsoldit]

so I used @treg_ai to index 3 months of twitter posts from all of my followers from X found 300+ opportunities to find people who can help me make gtm opensource

but when this post went viral i didnt build the infra to handle it (big lol)

so here's how i fixed everything.

[Jason Zhou @jasonzhou1993]

Been using @kylon_ai for a few month now,
It's almost like grok bot but built for team

Here is a quick walkthrough of how I use @treg_ai + Kylon to get our product cited by chatGPT - Comment if you want the prompt I use

[Treg @treg_ai]

@treg_ai is available in @kylon_ai from day 1!

[Treg @treg_ai]

Introducing OpenSource people-search for Claude

- Juts $0.0089/lead, no more $600 subscriptions
- #1 on People Search Bench
- 0% markup, Fully open source

Try it at https://t.co/D2eDbgiLu0

Repo link below 👇

[Akhilesh Mishra @livingdevops]

Kubernetes is easy until you run stateful workloads on it.
Your lead asks you to run Postgres on Kubernetes. Sounds reasonable. Your apps already run there.

So you deploy it as a Deployment. Postgres starts, you insert some rows. The pod restarts and your data is gone. The container filesystem dies with the pod.

So you use a PersistentVolume. Storage that lives outside the pod. The pod dies, a new one attaches the same disk, and your data survives.

But every volume needs someone to create it by hand. One DevOps engineer, four environments, six services. You become a bottleneck with a Slack handle.

So you use PVCs and StorageClasses. A developer asks for 10GB ReadWriteOnce, and Kubernetes provisions the disk automatically. No ticket, no AWS console.

Now you run 3 Postgres replicas on that PVC for high availability. All three write to the same disk with no coordination, and the data gets corrupted.

The problem is not storage. A Deployment treats every pod as identical. Postgres doesn't. Pod 0 is the primary, the others are replicas, and startup order matters.

So you use a StatefulSet. Stable names, stable identity, ordered startup. Pod 0 is always Pod 0. volumeClaimTemplates give every pod its own disk, so nothing is shared.

Your database is stable. Now you have dev, staging and prod. You copy the YAML three times and change the values. Six months later nobody knows why staging behaves differently from prod.

So you use Helm. One template, one values file per environment. No drift, and a bad upgrade is one rollback away.

Your team grows and everyone has cluster access. One team reads another team's secrets.

So you use RBAC. The payments team gets the payments namespace and the orders team gets orders. Nobody touches what isn't theirs.

Then one bad query eats every CPU on the node and starves everything running beside it.

So you use ResourceQuotas and LimitRanges. Quotas cap each namespace. LimitRanges give default limits to pods that never set their own. No single workload can take over the cluster.

Each concept exists because the previous one left something broken.

[Akhilesh Mishra @livingdevops]

Kubernetes is beautiful.

Every Concept Has a Story, you just don't know it yet.

In k8s, you run your app as a pod. It runs your container. Then it crashes, and nobody restarts it. It is just gone.

So you use a Deployment. One pod dies and another comes back. You want 3 running, it keeps 3 running.

Every pod gets a new IP when it restarts. Another service needs to talk to your app but the IPs keep changing. You cannot hardcode them at scale.

So you use a Service. One stable IP that always finds your pods using labels, not IPs. Pods die and come back. The Service does not care.

But now you have 10 services and 10 load balancers. Your cloud bill does not care that 6 of them handle almost no traffic.

So you use Ingress. One load balancer, all services behind it, smart routing. But Ingress is just rules and nobody executes them.

So you add an Ingress Controller. Nginx, Traefik, AWS Load Balancer Controller. Now the rules actually work.

Your app needs config so you hardcode it inside the container. Wrong database in staging. Wrong API key in production. You rebuild the image every time config changes.

So you use a ConfigMap. Config lives outside the container and gets injected at runtime. Same image runs in dev, staging and production with different configs.

But your database password is now sitting in a ConfigMap unencrypted. Anyone with basic kubectl access can read it. That is not a mistake. That is a security incident.

So you use a Secret. Sensitive data stored separately with its own access controls. Your image never sees it.

Some days 100 users, some days 10,000. You manually scale to 8 pods during the spike and watch them sit idle all night. You cannot babysit your cluster forever.

So you use HPA. CPU crosses 70 percent and pods are added automatically. Traffic drops and they scale back down. You are not woken up at 2am anymore.

But now your nodes are full and new pods sit in Pending state. HPA did its job. Your cluster had nowhere to put the pods.

So you use Karpenter. Pods stuck in Pending and a new node appears automatically. Load drops and the node is removed. You only pay for what you actually use.

One pod starts consuming 4GB of memory and nobody told Kubernetes it was not supposed to. It starves every other pod on that node and a cascade begins. One rogue pod with no limits takes down everything around it.

So you use Resource Requests and Limits. Requests tell Kubernetes the minimum your pod needs to be scheduled. Limits make sure no pod can steal from everything around it. Your cluster runs predictably.

[silentguy @silentguyy66]

Elon Musk's engineers showed what their workday looks like when the bots make every decision

In this 45-minute episode, Grok Bot engineers open their entire setup:

00:46 - Pen and Lauren from the Grok Bot team join
07:05 - three bots debate an idea in one group chat
22:10 - Lauren's bot books her whole multi-city trip alone
28:31 - agent prs automerge before she reads them
44:03 - "it's like hiring virtual employees"

Nobody shows the boring part

Lauren's chief of staff writes no code, it only picks which engineer bot gets the task

It all comes back to trust: how much are you willing to give a bot before you watch it work

More bots won't help if you still route every task yourself

Save this before XAI ships another Grok Bot update

[Shen Sean Chen @ShenSeanChen]

I made a video breaking down AI agent observability and evals, and ran the whole thing on open source.

If you build agents, sooner or later you watch one run and realise you have no idea what it just did, or what it cost you. That gap is what observability and evals are for. Least glamorous part of the stack, and the part you end up needing most.

🛠️ treg, OpenRouter for agent tools (4.2k): https://t.co/CQkPC6UZHj
🔑 Try treg: https://t.co/GqP6KAQwg7
🌟 Github Repo (2k): https://t.co/h1AyRL8jzR
💻 Portable memory: https://t.co/U2x120RmEI
☕ Join our community: https://t.co/NMcEekLZbL

Observability is a camera. It records what ran, how long, what it cost.

Eval is a referee. It decides whether the run was good enough to ship.

I ran it on @treg_ai for the tool calls and @waku_agent for the memory. treg is an OpenRouter for agent tool calls, so your agent carries one API key instead of one per tool.

Both open source, so you can read the code rather than take my word for it.

Full video in the first reply 👇

You Can Build Anything. You Can Learn Anything. 💪