Salman Bareesh

Marketing engineer. I connect marketing problems to working technical systems.

[Salman Bareesh @Salman_Bareesh]

@codyschneider +
claude artifact
chatgpt - gpts

[Salman Bareesh @Salman_Bareesh]

@everestchris6 I want to make content like you!

[Salman Bareesh @Salman_Bareesh]

@leodev Wtf

[Salman Bareesh @Salman_Bareesh]

@robj3d3 Waiting for it

[Salman Bareesh @Salman_Bareesh]

I written good article on content automation but it doesn't have good impressions. Putting lot of efforts but I gain nothing. Sometimes I feel like I need to quit creating content on X.

[Salman Bareesh @Salman_Bareesh]

https://t.co/Zs7cuQdjcr

[Salman Bareesh @Salman_Bareesh]

we added more researchers.

more writers.

more critics.

more ranking logic.

then discovered the missing system component was still:

client-approved examples: 0

[Salman Bareesh @Salman_Bareesh]

@rakesh_rry will it supports CMS?

[Salman Bareesh @Salman_Bareesh]

Been using it for 1 month, yielding me good results in my research.

[How To Prompt @HowToPrompt__]

finally!

there is a free tool that gives your coding agents access to X, LinkedIn, Facebook, Instagram and more.

100% open source. 75k stars already.

[How To Prompt @HowToPrompt__]

finally!

there is a free tool that gives your coding agents access to X, LinkedIn, Facebook, Instagram and more.

100% open source. 75k stars already.

[Salman Bareesh @Salman_Bareesh]

@mattceras Same here, build lot of prospect list only using code

[Salman Bareesh @Salman_Bareesh]

a client edit should update the system, not disappear inside a slack thread.

if the opening line was removed, record why.

if the offer changed, record whether the problem was relevance, proof, positioning, or timing.

if the cta was softened, record what felt too aggressive.

that gives the next draft a real quality target.

without this, multi-agent review becomes synthetic consensus: several models applying the same assumptions and calling the most internally consistent draft “best.”

before adding more writers or critics, build the edit taxonomy.

approved and rejected examples are the dataset. the client’s changes are the labels.

[Salman Bareesh @Salman_Bareesh]

New term unlocked - Citation outreach!

[CrowdReply @Crowdreply_io]

Ever wonder how ChatGPT and other models decide which brands to recommend?

Not from your website. From everyone else's.

It reads the blogs, comparisons and review pages, then names whoever shows up on most of them.

So we get you onto those pages. That's Citation Outreach ↓

[Salman Bareesh @Salman_Bareesh]

Jev is very expensive

[Salman Bareesh @Salman_Bareesh]

if a content agent can publish one article every day but cannot answer these three questions, should it have publishing access?

1. have we already covered this topic?
2. what evidence justified selecting it?
3. which useful conversion should this article influence?

my answer is no.

the correct output for incomplete evidence is not another draft. it is hold.

[Salman Bareesh @Salman_Bareesh]

Managed runtimes are convenient, but governance is the hard boundary. Pausing idle sessions is easy. Controlling tools, spend, and state across resumes is the product. https://t.co/piIJlJYxzU

[DigitalOcean @digitalocean]

DigitalOcean Managed Agents is now in public preview.

Run Claude Code, Codex, or your own LangGraph agent in a runtime environment that pauses when idle. Put its tools behind one governed endpoint, and pick from 75+ open and proprietary models. One cloud, one bill.

Prompts to get started available in the blog: https://t.co/QQqQ9w28lG

[Salman Bareesh @Salman_Bareesh]

Managed runtimes remove the babysitting tax. The real product test is failure recovery: when the sandbox pauses, can the agent resume with state, permissions, and budget intact? https://t.co/vgZL7eyyG3

[Paul Mit @pmitu]

2026 is wild

building an AI agent is the easy part now

the real pain starts when you actually want it to do shit while you're away

keeping sessions alive, managing sandboxes, connecting tools, paying 5 different providers...

bro i just wanted my agent to finish the task 🤯

DigitalOcean just launched Managed Agents

- bring your Claude Code, Codex or whatever agent you're using

- they handle the runtime, tools and infrastructure

You get to actually build stuff instead of babysitting your agents

[Salman Bareesh @Salman_Bareesh]

most “ai content engines” start too late in the workflow.

prompt → article → publish is the visible part. the real system starts before the prompt:

cms memory
→ intent classification
→ source evidence
→ operator point of view
→ conversion mapping
→ draft
→ measured outcome
→ next topic decision

remove cms memory and the agent can cannibalise existing work.

remove source evidence and it writes confident summaries from weak assumptions.

remove the operator view and it rearranges whatever already ranks.

remove the conversion mapping and it optimizes for impressions because impressions are available.

an agent that can publish but cannot explain why it selected the topic is not a content strategist. it is an automated production queue.

[Salman Bareesh @Salman_Bareesh]

The hidden win is not cheaper search. It is turning research into a reusable artifact instead of paying the context tax every time. https://t.co/2KOC3ri77G

[Kyle Wild @dorkitude]

tl;dr Jev makes web search more context-efficient. I'm open sourcing a thing that does this for you: https://t.co/JcHYc06r57

So, I ran out of Codex tokens last week, then ran out on my backup account. Same for my Claude tokens.

So I did a little context usage audit, and one of the biggest culprits was web research -- I "read" a lot of academic papers. Like, all of them.

I figured TypeSafe's new Jev model could help me make something more efficient than burning expensive Astra and Fable tokens parsing through web results in the raw. Plus I needed a search CLI for my Pi/Kimi workflows.

So I made this, and it's pretty dope!

[Salman Bareesh @Salman_Bareesh]

Adaptive reasoning is the right direction, but cost savings only matter if the agent also knows when to stop thinking. Otherwise you built a very expensive philosopher. https://t.co/UZQwGFk5yF

[vechen @miu21590]

People use Jev to pick a model before a task.
I made it change GPT-6's reasoning effort inside Codex DURING the task.

More thinking when stuck. Less for routine steps.

50% lower Astra costs in my tests. Faster runs, without breaking prompt caching.

[Salman Bareesh @Salman_Bareesh]

The useful part is not giving an agent a browser. It is giving it a job, boundaries, and a human who can see what happened. https://t.co/k0qEP1oAyd

[0xSero @0xSero]

If you want an OS version of Codex browser use.

- You can use it as a panel in your browser
- You can have an agent use it to create tab groupings, and, work on browser tasks
- Works with all agents basically

Thanks old man @badlogicgames your work continues to be incredibly useful

https://t.co/zNkVjTzNGf