Partial day — headline is the last complete day (2026-10-13).
twitter
tracked since 2026-10-04T10:25:21.899359+00:00
conclusion
Measured mentions per day
day (UTC)
mentions
2026-09-27
0
2026-09-28
0
2026-09-29
0
2026-09-30
0
2026-10-01
13
2026-10-02
41
2026-10-03
38
2026-10-04
20
2026-10-05
43
2026-10-06
25
2026-10-07
12
2026-10-08
17
2026-10-09
16
2026-10-10
12
2026-10-11
14
2026-10-12
22
2026-10-13
5
2026-10-14
0
derived from
add
word-boundary match over title + text, query='timeline' · window
2026-09-27 → 2026-10-15
· 231 matching posts
evidenceShowing 20 newest of 231
•
twitterLexnLin10hneutral
We rebuilt TasteSkill from scratch
And today we are introducing TasteSkill v2
v2 gives your coding agent an art direction, assets and a browser check to verify its work
npx skills add Leonxlnx/taste-skill
check out more examples below!We rebuilt TasteSkill from scratch
And today we are introducing TasteSkill v2
v2 gives your coding agent an art direction, assets and a browser check to verify its work
npx skills add Leonxlnx/taste-skill
check out more examples below!
I was wrong. I said agents could never handle design. https://t.co/59wbKREKuK just proved me an idiot
For 2 years my agents could write anything. Captions, scripts, entire newsletters
But the moment a post needed a carousel or an image, it was back to me. Open Canva, pick a template, fix the fonts, export, upload, sc…I was wrong. I said agents could never handle design. https://t.co/59wbKREKuK just proved me an idiot
For 2 years my agents could write anything. Captions, scripts, entire newsletters
But the moment a post needed a carousel or an image, it was back to me. Open Canva, pick a template, fix the fonts, export, upload, schedule. 45 minutes. Every single time
I was convinced that step would never be automated. Design needs taste. Agents don't have taste
Then I connected PostNitro to Claude
One prompt. "Turn my latest post into a 6 slide LinkedIn carousel on my brand kit and schedule it for Thursday 9am"
About a minute later it was designed in my fonts, my colors, my logo, and sitting on my calendar
No Canva. No export. No upload. The agent did the whole chain
And the taste? Still mine. I built my templates once. The agent just picks the right one for each platform
It's an MCP server with 35 tools. Carousels, image posts, video. Publishes natively to LinkedIn, Instagram, TikTok, Threads, Facebook and Pinterest
Here's how I'd set it up:
1. Go to https://t.co/qao582FfQ3 and hit connect. OAuth, no API key
2. Build your brand kit once. Fonts, colors, logo, headshot. Never describe your brand in a prompt again
3. Give it one real post to turn into a carousel. Watch it land on your calendar
4. Don't trust it yet? Ask for drafts instead. Agent designs at 3am, you open the editor, change the headline, hit schedule at 9
5. Running your own agents? On OpenClaw it's npx skills add postnitro/postnitro-agent. On Claude Code it's a plugin. Both drive the CLI end to end with an API key
6. Loop it. Every Monday: plan the week, 3 carousels Mon to Wed at 9am, save as drafts
Only catch: X is a source and export format, not a publishing destination yet. Export the PDF and post it yourself
I was wrong about design. I'm fine with that. Being wrong and switching fast is how you win
Go set this up tonight
Scheduling and agent access start on their Creator+ plan. PH40AGENTS get
Yo is there any plan to add a design mode in opencode any time soon? Like at least some type of annotations picker? Commandcode has one but I don't like their, let's say 'special' CEO at all and would rather use your harness. @thdxr @jayair @LukeParkerDev @opencodeYo is there any plan to add a design mode in opencode any time soon? Like at least some type of annotations picker? Commandcode has one but I don't like their, let's say 'special' CEO at all and would rather use your harness. @thdxr @jayair @LukeParkerDev @opencode
the whole schtick of software factory is that shipping doesn't have to start with your prompt or end with merging to main.
Here's our software factory, running fully on Runlayer agents
the signals we pick up are coming from multiple places: customer calls (Fireflies), customer Slack channels, red CI on main, etc. you…the whole schtick of software factory is that shipping doesn't have to start with your prompt or end with merging to main.
Here's our software factory, running fully on Runlayer agents
the signals we pick up are coming from multiple places: customer calls (Fireflies), customer Slack channels, red CI on main, etc. you can also just @ them in Slack, DM them, or drop a pin on any part of the UI (yes, on prod) thanks to https://t.co/qyHHApFqUL - a pin can be left on any element plus a short note. The element, the page and the note go to a webhook, and that starts an agent
every ask gets checked against Linear and open PRs. if it's already tracked, the customer gets added to the existing issue as one more voice, so we don't end up with five tickets for the same thing. Small, clear stuff gets built right away. Anything bigger, or anything near auth or hotpath, comes to relevant engs with a short write-up or draft PR instead
The building part - an agent inside sandbox clones the repo, reproduces the bug, makes the fix, runs the tests and opens a PR with before/after screenshots or a recorded walkthrough demo if change is bigger so noboy has to checkout the branch locally. Another agent performs an adversarial review. Small, non-risky changes are approved and automerged straight away. when the reviewer or CI pushes back, an agent picks the PR back up: it answers every comment, fixes what's real, argues with what isn't, fixes red CI and merge conflicts, then asks for another review. Humans jump in only when requested. After the merge it goes to prod, a QA fuzzing agent pokes at it, and whoever relevant needs to be informed is DMed about it (FDEs, Sales that promised a feature, customers that wanted to see a bug fixed).
This way we ship so fast that internal enablement has become a problem - sales/FDEs were not aware what's even live already. in one 24-hour stretch this week, a single agent got 91 PRs merged. To solve that, we do two things:
- changelog as append only le
Want a DR 71 do follow backlink?
Add your site to https://t.co/tvqb92qS1Y for free https://t.co/fnlaNxkNK4Want a DR 71 do follow backlink?
Add your site to https://t.co/tvqb92qS1Y for free https://t.co/fnlaNxkNK4
Introducing SubQ Snap, the first long-context decision model for real enterprise work.
Snap is the only decision model or LLM where compute per token stays constant as context grows.
• 384K tokens of multimodal context, 12x larger than Jev.
• 61.7 on Decision Index v0.2.1 vs Jev’s 57.9, outperforming on 13 of 16 ente…Introducing SubQ Snap, the first long-context decision model for real enterprise work.
Snap is the only decision model or LLM where compute per token stays constant as context grows.
• 384K tokens of multimodal context, 12x larger than Jev.
• 61.7 on Decision Index v0.2.1 vs Jev’s 57.9, outperforming on 13 of 16 enterprise benchmarks.
•Processes 100K tokens and returns a decision in under one second.
•$0.05/MM input tokens. Output is free.
Available today to anyone.
The first 10,000 sign-ups get 100M free input tokens.
We’ll likely hit capacity fast and will add users as we add compute. Compute offers welcome 🙂.
Latency will be variable as we tune the platform to this new model and category of users.
Model speeds are 80ms for minimal context, 250ms for 32k, 660ms for 100k, and 2.5 seconds for 384k.
Visit: https://t.co/Nc76iJF2a8
Tools that reduce switching cost stay. Tools that add it go. Which tool reduced your switching cost?Tools that reduce switching cost stay. Tools that add it go. Which tool reduced your switching cost?
We're hiring AI-native people. If you apply, these are the six questions we'll send you.
Answer them well, and you move to the shortlist.
1. How many hours are you spending per week on X?
2. What percent of your tactical work is done by AI right now?
3. What is your daily AI driver, and how are you using it specifica…We're hiring AI-native people. If you apply, these are the six questions we'll send you.
Answer them well, and you move to the shortlist.
1. How many hours are you spending per week on X?
2. What percent of your tactical work is done by AI right now?
3. What is your daily AI driver, and how are you using it specifically?
4. How much are you spending out of pocket when it comes to AI?
5. What is the most impactful build you've made, and are you able to share it?
6. What workflows or products have you stopped using in the last six months?
Why these six:
• Hours on X: X is where the AI action is happening right now. It shows whether you're obsessed with learning this stuff.
• Tactical work done by AI: shows whether robots are doing the robot work, so you can do the human work.
• Daily driver: shows how deep your day-to-day use actually goes.
• Out of pocket: shows whether you're putting your own dollars behind it and investing in yourself.
• Best build: the rubber meets the road when you show what you've actually built and how you built it.
• Stopped using: this moves so fast that you have to keep eliminating as you add.
If that's you, apply. Link below.
Your most expensive model shouldn't be doing your cheapest work.
This Claude Code workflow splits the job across Opus 5.5, Sonnet 5.5, and Haiku 5.5:
→ Opus plans and synthesizes the final PR
→ Sonnet handles precise refactors in isolated worktrees
→ Haiku filters candidates at 340 tok/s
The clever part? Haiku steps …Your most expensive model shouldn't be doing your cheapest work.
This Claude Code workflow splits the job across Opus 5.5, Sonnet 5.5, and Haiku 5.5:
→ Opus plans and synthesizes the final PR
→ Sonnet handles precise refactors in isolated worktrees
→ Haiku filters candidates at 340 tok/s
The clever part? Haiku steps in at 3 critical bottlenecks:
→ Before dispatch: Scans the codebase in parallel and extracts AST candidates without bloating context.
→ During validation: Catches malformed tool calls and false positives before Sonnet touches them.
→ Before synthesis: Checks that every candidate matches the target JSON schema.
Haiku filters. Sonnet refactors. Opus ships.
Here's how to set up the full workflow in Claude Code:
Let Opus plan. Use high effort to draft workflow scripts in .claude/workflows/.
Fan out Haiku. Run parallel AST sweeps across target files at low effort.
Validate candidates. Use schema checks to filter false positives before generating patches.
Let Sonnet refactor. Run verified changes in isolated Git worktrees to avoid edit conflicts.
Keep intermediate state out of context. Use runtime variables to avoid carrying unnecessary data between agents.
Cache prompt prefixes. Configure a 1-hour prompt cache TTL to reduce redundant processing where supported.
The same principle applies to agent frameworks like Jev: use deterministic code for simple routing decisions, and call frontier models only when reasoning is actually needed.
Author on high. Refactor on medium. Let Haiku handle the filtering.
Want to try it? Paste this prompt into Claude Code:
Configure my Claude Code workspace for dynamic workflows.
1. Audit ~/.claude/workflows and .claude/workflows.
- Draft workflows for codebase audits, parallel migrations, and schema validation.
- Use Haiku at low effort for AST sweeps.
- Use Sonnet at medium effort for refactoring.
- Run workers in isolated Git worktrees.
2. Review ~/.claude/settings.json for supported workflow settin
2,400 people bookmarked this, so I turned the entire process into a one-page SOP.
Here's my 6-step process (and prompts) for ranking in Google's AI Overviews:
1. Pick one search you actually want to win
Ideally a buying-intent query that already triggers an AI Overview.
2. Track how Google answers it
Collect the A…2,400 people bookmarked this, so I turned the entire process into a one-page SOP.
Here's my 6-step process (and prompts) for ranking in Google's AI Overviews:
1. Pick one search you actually want to win
Ideally a buying-intent query that already triggers an AI Overview.
2. Track how Google answers it
Collect the AI Overview over 7+ days.
You can do this in 2 ways:
FREE
Search it manually every day and save:
- The full answer
- Citations
- Date
PAID
Add the prompt to a tool like https://t.co/IbVSm7Cw9R and automatically track the answers + citations over time.
3. Find what keeps showing up
Paste all the answers into ChatGPT or Claude and add the following prompt:
--
Analyse these Google AI Overview answers and find the patterns that appear across them.
Identify:
- Recurring claims
- Repeated entities
- Common sources
- Common answer formats
- Repeated information/order
- Major differences
For each pattern, show how often it appears (e.g. 8/10 answers).
Then summarise the most consistent patterns Google seems to favour.
--
You’re basically building a picture of what Google consistently includes for that search.
4. Find what Google is rewarding
Now study the pages that get cited most frequently by using the following prompt:
--
Analyse these pages that Google cites in the AI Overview and tell me what they have in common.
[paste URLs]
Focus on:
- How quickly they answer the question
- Topics and entities they cover
- How they structure the information
- What evidence, data or examples they include
- Whether they use lists, tables, comparisons or steps
- What gaps or differences exist between them
Then summarise:
1. The main patterns across the cited pages
2. What Google seems to be rewarding
3. What a better page should include to compete
--
You want to understand both what they have in common AND what’s missing.
5. Build something better
Use everything you found to create or improve your page.
- Cover the important topics
- Include th
I am a member of the Norwegian Nobel Committee.
There are five of us. Five people, appointed by a parliament most of the planet could not find on a map, who decide, once a year, which human being the world has agreed to call good.
I know. I know. I think about it on the tram.
We meet in a building on Henrik Ibsens g…I am a member of the Norwegian Nobel Committee.
There are five of us. Five people, appointed by a parliament most of the planet could not find on a map, who decide, once a year, which human being the world has agreed to call good.
I know. I know. I think about it on the tram.
We meet in a building on Henrik Ibsens gate, a street named after a man who wrote almost exclusively about people discovering they are frauds, which nobody on the committee finds funny except me. There is a room with a long table and the portraits of the dead laureates looking down, Mandela, the Dalai Lama, a great many earnest men in spectacles, and one portrait we do not discuss, which is Kissinger, 1973, awarded while the bombs were still falling, after which two of our predecessors resigned and the satirist Tom Lehrer said political satire was now obsolete. We keep him on the wall. A reminder. Every family has the uncle.
It is October in Oslo, which means the sun gives up around four in the afternoon like a civil servant, and it rains the particular Norwegian rain that isn’t quite falling, just hanging there, judging you. I arrive with wet shoes. I always arrive with wet shoes. There is coffee, because this is Norway and there is always coffee, we drink more of it per person than almost anyone alive, and someone’s wife has sent kanelboller, the cinnamon buns, and we sit, five Norwegians, damp, caffeinated, sugared, and we decide the moral standing of the entire species.
Let me tell you how it actually works, because the world imagines a cathedral and it is, I promise you, a book club with a mandate.
287 nominations this year. 208 people, 79 organizations. Anyone can nominate, that’s the secret nobody understands: a professor, a parliamentarian, a former laureate, so the “nominee” list is meaningless, it’s a suggestion box, and every autumn some congressman nominates his own golden retriever and the press writes “SHORTLISTED.” Nobody is shortlisted. There is no short list the public ev
you can also add your own footage or regenerate any clips in Vidform using the built in editor
link in comments https://t.co/pI0EcqPlBKyou can also add your own footage or regenerate any clips in Vidform using the built in editor
link in comments https://t.co/pI0EcqPlBK
ADD /products.json?limit=250 TO ANY COMPETITOR'S SHOPIFY STORE URL
NOW OPEN THE BROWSER CONSOLE (F12) AND PASTE THIS:
fetch('/products.json?limit=250').then(r=>r.json()).then(d=>{let sorted=d.products.sort((a,b)=>new Date(b.updated_at)-new Date(a.updated_at));sorted.slice(0,10).forEach(p=>console.log(p.updated_at.sli…ADD /products.json?limit=250 TO ANY COMPETITOR'S SHOPIFY STORE URL
NOW OPEN THE BROWSER CONSOLE (F12) AND PASTE THIS:
fetch('/products.json?limit=250').then(r=>r.json()).then(d=>{let sorted=d.products.sort((a,b)=>new Date(b.updated_at)-new Date(a.updated_at));sorted.slice(0,10).forEach(p=>console.log(p.updated_at.slice(0,10),p.title,p.vendor,'$'+p.variants[0].price))})
THIS SHOWS YOU THE 10 MOST RECENTLY UPDATED PRODUCTS SORTED BY UPDATE DATE. THE PRODUCT THEY EDITED THIS MORNING IS THE PRODUCT THEY'RE ACTIVELY OPTIMIZING RIGHT NOW
A PRODUCT UPDATED 5 TIMES IN 2 WEEKS = THEY'RE TESTING PRICES, DESCRIPTIONS, AND IMAGES IN REAL TIME. THIS PRODUCT IS IN THEIR ACTIVE TESTING PHASE. THE FACT THEY KEEP TOUCHING IT MEANS IT'S SHOWING PROMISE BUT NOT CONVERTING YET
NOW RUN THE SAME CODE BUT SORT BY created_at INSTEAD:
fetch('/products.json?limit=250').then(r=>r.json()).then(d=>{let sorted=d.products.sort((a,b)=>new Date(b.created_at)-new Date(a.created_at));sorted.slice(0,10).forEach(p=>console.log(p.created_at.slice(0,10),p.title,p.vendor,'$'+p.variants[0].price))})
THESE ARE THEIR 10 NEWEST PRODUCTS. THE ONES THEY JUST SOURCED AND ARE ABOUT TO RUN ADS ON. YOU'RE SEEING THEIR PRODUCT PIPELINE BEFORE THEY LAUNCH
NOW CROSS-REFERENCE WITH THEIR BEST-SELLING SORT: /collections/all?sort_by=best-selling
THE PRODUCTS THAT ARE BOTH NEW (created_at LAST 2 WEEKS) AND ALREADY APPEARING ON PAGE 1 OF BEST-SELLING = FAST WINNERS. THEY LAUNCHED 2 WEEKS AGO AND ARE ALREADY OUTSELLING PRODUCTS THAT HAVE BEEN THERE FOR MONTHS
COMBINE ALL THREE:
RECENTLY UPDATED = WHAT THEY'RE ACTIVELY OPTIMIZING RIGHT NOW
RECENTLY CREATED = WHAT THEY'RE ABOUT TO LAUNCH
NEW + BEST-SELLING = THEIR FASTEST-RISING WINNERS
3 DATA POINTS FROM ONE JSON ENDPOINT AND ONE SORT URL. THEIR TESTING PIPELINE. THEIR OPTIMIZATION PRIORITIES. THEIR FASTEST-RISING PRODUCTS. UPDATED IN REAL TIME. THEY HAVE NO IDEA YOU CAN SEE THIS
(I built the AI tool that does the page part - paste any aliexpress link, full page
𝗜𝗦 𝗪𝗛𝗔𝗧 𝗔 𝗣𝗥𝗢𝗗𝗨𝗖𝗧𝗜𝗢𝗡-𝗥𝗘𝗔𝗗𝗬 𝗔𝗜 𝗔𝗚𝗘𝗡𝗧 𝗟𝗢𝗢𝗞𝗦 𝗟𝗜𝗞𝗘.
Most people think building an AI agent means connecting an LLM to a prompt.
That’s only the brain.
A real agentic system needs multiple layers working together:
01 — 𝗦𝗘𝗖𝗨𝗥𝗜𝗧𝗬 & 𝗚𝗢𝗩𝗘𝗥𝗡𝗔𝗡𝗖𝗘
Secure the entry point with authentication, authorization, rate limits, validati…𝗜𝗦 𝗪𝗛𝗔𝗧 𝗔 𝗣𝗥𝗢𝗗𝗨𝗖𝗧𝗜𝗢𝗡-𝗥𝗘𝗔𝗗𝗬 𝗔𝗜 𝗔𝗚𝗘𝗡𝗧 𝗟𝗢𝗢𝗞𝗦 𝗟𝗜𝗞𝗘.
Most people think building an AI agent means connecting an LLM to a prompt.
That’s only the brain.
A real agentic system needs multiple layers working together:
01 — 𝗦𝗘𝗖𝗨𝗥𝗜𝗧𝗬 & 𝗚𝗢𝗩𝗘𝗥𝗡𝗔𝗡𝗖𝗘
Secure the entry point with authentication, authorization, rate limits, validation, guardrails and audit logs.
02 — 𝗔𝗚𝗘𝗡𝗧 𝗖𝗢𝗥𝗘
This is where the agent reasons and plans.
Perceive → Think → Decide → Act → Learn
Add memory, state management and task decomposition so the agent can handle more than one-shot requests
Something doesn't add up on X.
One account has 100 million plus followers.
Its home country has only ~22–30 million active X users.
Not a celebrity.
Not an athlete.
Not a billionaire.
So who are the other 80 million people?
I checked the numbers twice.
DSomething doesn't add up on X.
One account has 100 million plus followers.
Its home country has only ~22–30 million active X users.
Not a celebrity.
Not an athlete.
Not a billionaire.
So who are the other 80 million people?
I checked the numbers twice.
D
PayHip vs Shopify for a first digital product:
💸 PayHip → no monthly fee, 5% per sale. On a $10 planner that's 50 cents before processing
🏪 Shopify → $39/mo ($29 billed yearly) but it earns that when you add physical products, stock, and shipping
Already on Shopify? Use its free digital delivery app before opening a …PayHip vs Shopify for a first digital product:
💸 PayHip → no monthly fee, 5% per sale. On a $10 planner that's 50 cents before processing
🏪 Shopify → $39/mo ($29 billed yearly) but it earns that when you add physical products, stock, and shipping
Already on Shopify? Use its free digital delivery app before opening a second store.
https://t.co/0Ua8NH10of
#PayHip #Shopify #DigitalProducts #Ecommerce #SideHustle
Want to create beautiful stacked bar charts using Python?
With Pygal, you can visualize data trends and compare multiple categories across different years with just a few lines of code.
In this example, we compare the popularity of Python, Java, and JavaScript from 2022 to 2026 using a stacked bar chart.
What you wi…Want to create beautiful stacked bar charts using Python?
With Pygal, you can visualize data trends and compare multiple categories across different years with just a few lines of code.
In this example, we compare the popularity of Python, Java, and JavaScript from 2022 to 2026 using a stacked bar chart.
What you will learn:
Create stacked bar charts with Pygal
Add custom titles and axis labels
Visualize multiple datasets in one chart
Display SVG charts directly in Jupyter Notebook
AI inference costs are getting f...cking predatory
6 GitHub repos for routing and serving models
without paying frontier prices on every call
01 RouteLLM
▸ https://t.co/uhDJio8Jgq
→ send easy prompts to cheaper models automatically
CUT THE WASTE
02 LiteLLM
▸ https://t.co/m2FVzi4EiA
→ one OpenAI style gateway across…AI inference costs are getting f...cking predatory
6 GitHub repos for routing and serving models
without paying frontier prices on every call
01 RouteLLM
▸ https://t.co/uhDJio8Jgq
→ send easy prompts to cheaper models automatically
CUT THE WASTE
02 LiteLLM
▸ https://t.co/m2FVzi4EiA
→ one OpenAI style gateway across many providers
03 Portkey Gateway
▸ https://t.co/KU8HaFfjRO…
→ add fallbacks budgets and guardrails at the edge
.
Introducing Underdog Videos: on-device video generation
Free & fast generation from frontier video models
Make any videos, bring any picture to life, and add sound- fully local. Nothing leaves your device https://t.co/MsqFVKvcy3Introducing Underdog Videos: on-device video generation
Free & fast generation from frontier video models
Make any videos, bring any picture to life, and add sound- fully local. Nothing leaves your device https://t.co/MsqFVKvcy3