Minh Ha (@_mihado) header
Minh Ha (@_mihado) profile picture

There I am, therefore I think? https://xitter.cf/t.co/ZLbCaSvDji

Wellington
Joined March 2009
Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)Photo by Minh Ha (@_mihado)
It feels like we're all trapped in the endless cycle of do more, win more, more more more. Gonna try building an app to lowkey weave mindfulness into daily life. Let's see where these threads land.
1
1
6
548
The clankers are starting to take over all CPUs around me ...
78
When the easy stuff gets easier, only the hard stuff remains. To stay sane, better get used to doing hard things.
1
60
Working with the clankers is mind boggling. Very smart architect when being a co-driver. But occasionally will behave like a paranoid consultant and stack 12 guard rails on the most harmless change. Bikeshedding AGI ...
44
Thanks @tan_stack for the new ActivityStore
TanStack AI 1.0 is officially stable! 🎉 Build chat, voice, image generation, agents, and harnesses into your apps, with your choice of models, UI framework, and infrastructure. Your AI, your way. tanstack.com/ai
1
36
While Opus & DeepSeek are clanking on features & security, maybe it's about time I hire @bot to clank on some other tangents 🤔
32
About time to do KMS in local dev too, then it'll be seamless prod deployment github.com/floci-io/floci
1
36
Maybe the next evolution of the web is Workflows as a Service 🤔
25
Someone Is Impersonating FluidVoice, and we genuinely need your help Today, I'm sad to say that the system is broken. For the past year, we've been building FluidVoice, a free dictation app now used by 50,000+ people across the globe. It helps people with disabilities communicate and people with ADHD stay focused. We kept it free because we believe tools like this shouldn't be locked behind a paywall. We've been building it on our own time, with no pay. We never expected anything in return. We just wanted to make something useful, earn people's trust, and do right by our users. And for months now, a fraudulent website has been impersonating us. Today, it's the SECOND result on Google when someone searches for FluidVoice. They even claim it as "OFFICIAL DOWNLOAD", lol. (Please do not click on it!) The real FluidVoice is only at - altic.dev/fluid - fluidvoice.com We've tried everything we can think of to take it down. But nothing has worked. @Dynadot, the registrar, sends us a canned reply and tells us to go file a legal dispute. @HostGator, the host, wants us to pay $1500 to take legal action? @Cloudflare doesn't respond at all. NOTHING has worked. No one is helping put an end to this. So here we are. We give our work away for free, and now we're being told we have to pay to stop someone from pretending to be us. It's ridiculously easy for shameless opportunists to steal someone's identity, put up a fake website, impersonate something people have poured their heart into building, and get away with it. How is it so easy to impersonate someone, and this hard to stop it? Why does the burden fall on the people being impersonated instead of the people enabling it? Doesn't this feel like it should be illegal? We should be spending our time making FluidVoice better for people who rely on it. Instead, we're spending it fighting people who are pretending to be us. I'm genuinely tired of this. If you believe this is wrong, please repost. I just want this to reach someone who can help us put an end to it. It shouldn't be this hard to protect something you've worked so hard to build. Is there anyone at all who can help?
20
35
7
132
7,102
In university, I preferred horror movies because they jump-started the brain. But if The Ring had been like this, don’t think I could have studied at all 😹😹😹
贞子从电视里跑出来,夫妻不但不害怕,还把她当女儿养!!!
33
getting up & walking around do wonder to design block.
1
28
Over two weeks, Prime Agent orchestrated a swarm of over 2,000 agents to rewrite itself in Rust. With 10,000+ sandboxes, 200B+ GLM-5.3 tokens, and 16,000 agent-to-agent messages, this was our largest test yet of Prime Agent’s multi-agent capabilities and our infrastructure powering these agent swarms. Prime Agent now reaches usable input ~13× faster and uses 83% less startup memory than before.
186
207
146
2,842
565,805
If you learn GTM Engineering, you will never be unemployed again. Stage 1: GTM Data Architecture and Reverse ETL - Learn: Snowflake/BigQuery, Hightouch/Census, identity resolution, API webhooks. - Practice: Build a pipeline that syncs raw Stripe billing data and product usage directly into HubSpot/Salesforce custom objects. - Why: GTM without a single, code-level source of truth is just guessing. Stage 2: Signal and Intent Pipelines - Learn: Website deanonymization (Clearbit), intent data (Bombora), event streaming. - Practice: Build a webhook router that instantly alerts a Slack channel when a target enterprise account visits your pricing page 3 times in one week. - Why: In B2B, timing beats the pitch. You must strike while the intent is hot. Stage 3: AI SDRs and Hyper-Personalization - Learn: Clay, Apollo APIs, LLM data enrichment, web scraping. - Practice: Build an AI agent that scrapes a prospect's recent GitHub commits or podcasts and writes a highly specific, 1:1 cold email. - Why: Generic outbound is dead. AI makes infinite personalization mathematically possible. Stage 4: Product-Led Growth (PLG) Telemetry - Learn: PQL (Product Qualified Lead) scoring, event tracking (Segment/Rudderstack), feature adoption curves. - Practice: Write a script that triggers a high-priority sales alert the second a free-tier user hits a specific "aha moment" API threshold. - Why: The product is your best salesperson. Hand-raisers convert 5x higher than cold outbound. Stage 5: Usage-Based Pricing (UBP) and Monetization - Learn: Stripe Metered Billing, credit systems, value-metric alignment, API metering. - Practice: Design and implement a pricing architecture that charges by "agent executions" or "tokens processed" instead of flat "per-seat" licenses. - Why: SaaS pricing has permanently shifted from charging for access to charging for outcomes. Stage 6: Revenue Operations (RevOps) and Attribution - Learn: Multi-touch attribution, funnel leakage analysis, SQL, cohort analysis. - Practice: Build a dashboard proving exactly which combination of LinkedIn ads, webinars and cold emails sourced your closed-won enterprise deals. - Why: If you cannot mathematically prove ROI, the CMO will cut your budget. Stage 7: AI Sales Enablement and Coaching - Learn: Gong/Chorus APIs, transcript summarization, vector search on sales calls. - Practice: Build a tool that analyzes 100 lost sales calls and auto-generates an updated "objection handling" battle card for the sales team. - Why: Reps forget 70% of their training. AI reinforcement loops fix the leaky bucket. Stage 8: Predictive Churn and Expansion (NRR) - Learn: Health scoring algorithms, Net Revenue Retention (NRR) math, usage drop-off alerts. - Practice: Build a "churn risk" model that flags accounts with dropping API usage 45 days before their renewal date. - Why: In the AI era, switching costs are low. Retention and expansion are the only real growth. Stage 9: Integration-Led Growth (Ecosystem GTM) - Learn: App marketplaces, OAuth flows, Zapier/Make templates, co-selling motions. - Practice: Build a native integration for a major platform (like Slack or Salesforce) and optimize its marketplace listing to drive inbound leads. - Why: Buyers trust their existing stack more than your cold email. Ecosystems are the ultimate distribution. Stage 10: Dark Social and Community Attribution - Learn: Untrackable attribution, Discord/Slack community ops, developer advocacy metrics. - Practice: Launch a niche technical community and build a survey-based model tracking how "word of mouth" correlates to enterprise pipeline. - Why: B2B buyers buy from peers in private Slack groups not from vendor whitepapers. Stage 11: The Automated QBR (Quarterly Business Review) - Learn: Data visualization, automated reporting, ROI calculation frameworks. - Practice: Build a script that pulls customer usage data and auto-generates a custom, branded ROI slide deck for the Customer Success team. - Why: CSMs should be strategic advisors not manual data gatherers. Stage 12: Public GTM Teardowns and Portfolio - Learn: Pricing psychology, funnel teardowns, public writing, benchmark analysis. - Practice: Publish 3 deep-dives on how top AI companies structure their pricing pages, onboarding flows, and credit systems. - Why: GTM leaders and Revenue Architects are hired for their strategic judgment, not just their SQL skills. Code builds the engine. GTM builds the fuel. The modern GTM Engineer builds the nervous system for global revenue and autonomous sales agents. Bookmark and Repost!
36
49
7
487
45,076
Anyone out there shipping enterprise level works with DeepSeek v4.1? It needs lots of design thinking in loop but that’s also where the fun is.
21
When time is in deficit, scheduling anything is hard. Might as well keep doing hard things.
1
2
27
Finally got a legal & compliance RAG working reliably with @DBOS_Inc & @weaviate_io Now I need to learn proper eval & LORA.
1
1
1
84
Mono repo, microservices are solved with agents. Architecture, deep module, data boundaries are solved by human. Let’s try to keep it that way. That’s where the fun is.
20
a @Cloudflare tunnel, a @traefik container, fastest way to get reliable https for dev. Good time.
20
This is still the most enjoyable way to code with AI. You just need a model smart enough, cheap enough, and fast enough to riff with until values and understanding come out.
i started lode coding in early 2024 and casually punted it on here as “fleek coding”. after a year i realised it was worth sharing so i renamed it to lode coding and published my first article on it in early 2025. some seem to see the markdown file hierarchy and assume it’s just another human-curated knowledge base and then say “oh this is like <x>” and then move on. <x> is usually younger, human intensive, and requires additional tooling. lode coding is frictionless and tool-less. all you need is one prompt. you never write a single markdown file. you converse with the model, sharing your design and implementation goals, domain knowledge, and practices with it. the lode is 100% maintained by the model. all you do is work with it like it’s a new team member… … you focus on building software… … until the magic happens. once the lode is dense enough the model can receive instructions like you might give to a junior/mid colleague. your prompts are shorter and more domain and design focused. the word “lode” is deliberate. in mining a “lode” is a seam of valuable material. in our case that material is domain, design, and practices knowledge. knowledge that maps directly to the implementation. it’s a form of cognition for the model, seeded from what you told it about the problem, the design, your practices, etc. it’s highly coupled to the project at hand. lode coding is also more than just a self improving knowledge base, it’s a method. the lode is the foundation that allows you to effortlessly follow the method. the primary goal: your cognitive ownership. lode coding is the remedy for PR exhaustion and the antithesis of unattended loops which trade hysteria for cognitive damage, cognitive debt, and an irresponsible lack of ownership. throughout my journey with lode coding i set out to achieve one thing and one thing only: to leverage the speed of AI development while ensuring i can continue to practice my craft. i think i’ve achieved that. lode coding is in permanent evolution and i will share more about the most recent updates later this weekend. fjzeit.github.io/lode
1
48