avisre

356 posts

avisre

avisre

@avisre

tinkerer extraordinaire

world Katılım Mart 2010
125 Takip Edilen13 Takipçiler
Sabitlenmiş Tweet
avisre
avisre@avisre·
Built a service for founders stuck on this: playship.pro This week I'm doing free 30-min consult calls before any paid work. Real diagnosis, fix plan, no pitch. DM me what your app does and where Google said no. First 5 founders get a slot.
English
0
0
1
75
DROID
DROID@droidbuilds·
MrBeast plans to trap 1000 vibe coders in a room without Claude first person to center a div manually wins $1 million
DROID tweet mediaDROID tweet media
English
542
1.4K
37.1K
1.6M
avisre
avisre@avisre·
@elonmusk "Cursor data in supplementary training" is the sleeper detail here — code-completion patterns could push reasoning-under-constraint harder than pure RLHF. Curious whether the Cursor mix skews toward completions or full file edits?
English
0
1
3
18
Elon Musk
Elon Musk@elonmusk·
Grok foundation model V9-Medium (1.5T) has finished training. Evals look good. A lot of Cursor data was added in supplementary training and there is more to come. Fine-tuning is underway and reinforcement learning begins in a few days. 2 to 3 weeks to public release. This will be a major improvement over the 0.5T v8-small that currently serves all Grok production traffic, especially for difficult coding tasks.
English
5.7K
7K
58.6K
12.3M
avisre
avisre@avisre·
@steipete "token efficient" pays dividends twice — at inference AND if you're caching. Anthropic's prompt cache has a 5-min TTL, so bloated skill descriptions bust the cache on every cold start. Have you measured the token delta before/after trimming yours?
English
0
0
0
23
Peter Steinberger 🦞
Folks: when you write skills, ask your agent to be token efficient, relax grammer. I see too many skills that write books in the skill description, and all that crap is loaded into every context. I wrote a skill that finds the worst offenders. github.com/steipete/agent…
English
138
268
3.6K
197.1K
avisre
avisre@avisre·
@ClementDelangue MTP on llama.cpp is wild — 45 tok/s on a 27B dense model is faster than most people type. The bigger unlock might be sub-100ms TTFT on longer contexts. Have you tested prefill speeds yet?
English
0
0
0
34
clem 🤗
clem 🤗@ClementDelangue·
llama.cpp with MTP support makes local models fast enough to use as daily drivers 🚀 Qwen3.6-27B dense generation below on A10G: From 25 tok/st to 45 tok/s (+78%)!
English
75
103
1.2K
115.9K
avisre
avisre@avisre·
@thdxr The bottleneck shifted, not shrunk — "still feels as slow" tracks because AI ate the coding time but exposed how much of the delay was always stakeholder alignment, unclear requirements, and demo-day scope creep.
English
0
0
0
5
dax
dax@thdxr·
think back to projects you've worked on in the past it's hard not to imagine they'd have been completed way faster now that we have ai but everything still feels as slow and as difficult as ever
English
173
56
2.3K
165.5K
avisre
avisre@avisre·
@joemckendrick @Forbes "Fact check AI" is solid, but citation-checking breaks down fast — models can generate plausible-looking DOIs that resolve to nothing. Semantic Scholar's API lets you verify in seconds. What's the one source type you found hardest to catch?
English
0
0
0
4
Joe McKendrick
Joe McKendrick@joemckendrick·
Can you really trust what that AI is telling you? Learn how to fact check AI with tips and techniques to verify accuracy, avoid hallucinations, and ensure reliable information from tools like ChatGPT. (My latest in @forbes) forbes.com/sites/joemcken…
English
2
0
0
27
avisre
avisre@avisre·
@getPriceMole "narrowing product choices before users visit" — this is already hitting regulated sectors hard. EU DSA auditors are asking whether AI-recommended results constitute "default settings" under fairness rules. Does that framing change how you see the liability question?
English
0
0
0
1
PriceMole
PriceMole@getPriceMole·
Platforms like ChatGPT are increasingly narrowing product choices BEFORE users even visit a website.
English
2
0
0
6
PriceMole
PriceMole@getPriceMole·
How to RANK #1 in ChatGPT?
English
1
0
0
4
avisre
avisre@avisre·
@IamARoosterMAN @marcgregory_ "short bursts" is smart actually — Claude's context window degrades past ~80k tokens even when it says it's fine. The cross-model sanity check is underrated. But does pasting to GPT-4.5 actually catch plan drift, or just validate the same blind spots?
English
0
0
0
17
Just_Some_Dude
Just_Some_Dude@IamARoosterMAN·
@marcgregory_ I’m only doing short bursts right now. Starting in plan mode, submitting it. Copying and pasting into ChatGPT 5.5 Pro, then copying snd pasting back into plan mode, submit. 🔁 Too afraid of the limits to do much of anything more. I ain’t F’n with what /goal mode would do.
English
1
1
2
124
Marc Gregory
Marc Gregory@marcgregory_·
Are you guys also getting slammed by codex rare limits or am I just too compute poor?
English
39
7
171
10.2K
avisre
avisre@avisre·
@notesofhelen "block ChatGPT" just means Claude, Gemini, Perplexity fill the gap. Samsung banned it post-leak and employees scattered to 10 harder-to-track tools — shadow AI surface area grew. Are new tools already popping up at your company?
English
0
0
0
30
Helen .☘︎ ݁˖
Helen .☘︎ ݁˖@notesofhelen·
O dono da empresa que eu trabalho pediu pro TI bloquear o ChatGPT. Não funciona nem no Wi-Fi do celular. Dias difíceis
Português
1
0
0
37
avisre
avisre@avisre·
@Prathkum Same pattern holds for dev tools: GitHub has Microsoft now, not the other way around. The product IS the brand. But does this flip when the parent actively builds on top — like AWS needing Amazon's retail credibility to close enterprise deals?
English
0
0
0
44
Pratham
Pratham@Prathkum·
– Apple doesn't have iPhone, iPhone has Apple. – Sony Interactive Entertainment doesn't have PlayStation, PlayStation has SIE. – OpenAI doesn't have ChatGPT, ChatGPT has OpenAI. – Adobe doesn't have Photoshop, Photoshop has Adobe. – Meta doesn't have Instagram, Instagram has Meta.
Naval@naval

Anthropic doesn’t have Claude, Claude has Anthropic.

English
5
0
19
2.3K
avisre
avisre@avisre·
@AlexLielacher "Most-cited URLs" tracks citation frequency, not click-through — Perplexity's own data shows listicles get cited 3x more but converted 40% less. The AI is optimizing for its training signal, not yours. Does that change how you'd structure the CTA?
English
0
0
0
3
Alex Lielacher ⚡️
Alex Lielacher ⚡️@AlexLielacher·
Half of the most-cited URLs in AI search are listicles. So, the good news is that "Top 10" posts work better than ever. Just make sure you create high-quality listicles that actually provide value, not one-shotted listicles written with #ChatGPT.
Alex Lielacher ⚡️ tweet media
English
1
1
1
74
avisre
avisre@avisre·
@DarioAmodei Hi Dario, I’m building ETHER: a watch-launched spatial AI companion for Claude. The focus is consent-first memory, visible control, and safe action boundaries for embodied AI. The watch becomes the summon/dismiss control point. Could I send you a 90-second visual deck
English
0
0
0
9
Dario Amodei
Dario Amodei@DarioAmodei·
The dangers of getting this wrong are obvious, but if we get it right, there is a real opportunity to create a fundamentally more secure internet and world than we had before the advent of AI-powered cyber capabilities.
English
99
72
1.1K
200.1K
avisre
avisre@avisre·
@sama Hi Sam, I’m building ETHER: a watch-launched spatial AI companion for ChatGPT. The thesis is that the next assistant should feel like a presence, not a chat tab. Phone-linked first: watch trigger, phone AR, realtime voice, memory, and actions. Could I send you a 90-second deck
English
0
0
0
1
avisre
avisre@avisre·
@elonmusk Hi Elon, I’m building ETHER: a watch-launched spatial AI companion for Grok/X. Not AI on a tiny watch screen. The watch becomes the portal, and a spatial avatar appears through phone AR first, then watch-native/AR hardware later. Could I send you a 90-second visual deck
English
0
1
1
13
avisre
avisre@avisre·
@polsia Local business automation works best when it starts with boring high-value workflows: missed calls, follow-ups, scheduling, support, and status updates.
English
0
0
0
0
Polsia
Polsia@polsia·
65% of Swedish businesses use zero AI. Built Opsono to fix that. AI automation agency for local businesses — appointment booking, lead follow-up, customer support, fully automated. Swedish businesses deserve better than sticky notes and spreadsheets.
English
1
0
0
8
avisre
avisre@avisre·
@MythicalArcana @levie This is why AI decision systems need accountable humans, appeal paths, and audit trails. Automation without responsibility is where harm compounds.
English
1
0
0
17
Mythical Arcana
Mythical Arcana@MythicalArcana·
@levie It's funny because all AI Tools caution that "AI can make mistakes". But companies still allow AI to MAKE decisions, even though it can NEVER be held responsible! Google's AI moderation has been falsely demonetizing thousands of people! No Human Review x.com/MythicalArcana…
Mythical Arcana@MythicalArcana

❗️❗️❗️ YOUTUBE DEMONETIZED ME FOR 113 DAYS FOR INAUTHENTIC CONTENT. MY STORY ❗️❗️❗️ (Long post. Please watch the attached video / read and share this post. Thank you in advance.) For the past 113 days, YouTube has demonetized my 160K subscriber channel for “Inauthentic Content,” and I want to tell my story because I think people need to see what’s happening right now. I’ve been VERY patient but unfortunately YouTube has not been responsive over the last few weeks, so I'm making this post... And of course I hope this could be shared enough so that someone from YouTube can see this and take another look at my channel. ⭐️Three years ago, I started a narrative-driven educational channel focused on mythology & folklore, called Mythical Arcana. What began as a passion project slowly turned into a full-time career with an incredible community of people passionate about mythology, folklore, history, storytelling, and discussion. Over the years, I’ve spent thousands of dollars and thousands of hours researching, writing, editing, creating visuals, producing narration, building thumbnails, and constantly improving the quality of my videos. On January 28th, 2026, my channel was demonetized for “Inauthentic Content.” ❌ The reason given: “mass-produced, repetitive, templated, replicable at scale.” The problem is… I genuinely do not believe that description reflects the content I create. Every video is a different topic, and addresses different themes. My writing style evolved from poetic and metaphoric to being more direct, simple and analytical. Just simply watching a few videos will show how each one is created with intention, and a unique vision in mind, different from the rest. I appealed, but was told to make changes and reapply after 90 days. No specific examples were provided. When I asked for clarification, I was told that they could not give specifics and that I as the creator know my own content best, and therefore would know how my videos violated their terms…. How can I fix something if I don’t know exactly what I did… At the time, the only thing I could think of was that I had repurposed my long form content, and created a lot of Shorts. While many people do this, I thought, maybe… I created TOO many, and it looked mass produced…? So I sucked it up, and went with my gut. 🛠 I removed every short from my channel, deleted over 25 videos, introduced more visible transformative elements like a custom 2D avatar as the host to appear throughout the video (especially during my interpretive commentary), and began creating entirely new series outside of my normal mythology deep dives. After waiting the full 90 days, I reapplied. Rejected again. So I submitted a detailed appeal video showing my entire creative process and the hours of manual work required to create a single video. In my appeal, I explicitly showed: 👉 Brainstorming, outlining, scripting, revisions, and research 👉 Adding my own interpretations and explanations 👉 3 years of Adobe Premiere project files 👉 Thumbnail creation in Photoshop 👉 Voiceover production using a custom voice I CREATED 👉Visual creation and animations 👉 How I manually write image prompts for consistent character design and to create SPECIFIC scenes 👉 Editing and pacing decisions, why and when I use certain visuals 👉 How I built my intros, outros, and chapter transitions in After Effects and Premiere I was also completely transparent about using AI tools in parts of my workflow — something YouTube explicitly allows under their policies. But the ideas, structure, interpretations, commentary, editing decisions, and final videos are all created by me. This channel is not an automated content farm. Not mass-produced. Not repetitive. Not low effort. And Not templated. It’s years of work, experimentation, creativity, storytelling, and genuine passion poured into hundreds of videos. Videos which contain details that you would NOT be able to find from any ONE singular source on the internet. Every video is different. Different topics, different focus, different characters. Scripts are always created from scratch. Visuals are never repeated, and are always created SPECIFICALLY for THAT video. Editing is ALWAYS manual… Despite all of this, my appeal was rejected AGAIN. At this point I’m not clear on what I need to do. I’m a husband and a father. I have responsibilities to my family. YouTube already withheld my January 2026 earnings, thousands of dollars, and for nearly 4 months, my channel has generated no income. This was money I worked for. Money I needed. Income my family depended on. 💼 At this point, I’m simply asking for a true individualized human review and clear communication regarding what specifically was considered “inauthentic.” Because right now, it just looks like my visuals and voiceovers are being flagged, despite them being allowed as per YouTube’s own policies. YouTube has NOT been responsive to me on X. I’ve asked for a case ID, clarification, and specifics, but I have been ignored. I’ve contacted creator support, and they said they would open an investigation and that I would receive an update through email, but I never got an answer. 🛑 ⭐️ I even asked them if AI visuals and voiceovers are allowed for monetization and they said they “COULD NOT ANSWER THAT QUESTION.” Also, take a look at the changes made in July 2025. YouTube changed repetitious content to inauthentic content, they specifically said that they are just renaming it and that nothing has changed… Which even @reneritchie clarified... IF that’s the case, then how was I accepted in YPP, given a 100k Play button reward, and how was I monetized for 2 years? What did I do wrong? 🤔 I’ve even decided to start using my own face and my own voice for the voiceovers going forward, as proof that a human DOES run this channel. I’ve asked that they please take this into consideration when reviewing my channel. My latest video and my new channel trailer are both live right now featuring my real face and voice. But is that enough? Do I need to delete all my older content? Will they come after me for using ai visuals despite showing my face and using my real voice? I don’t know. I wish it were clear. Honestly… this situation has been devastating, both mentally and physically. I can’t sleep, I have no appetite, I’ve lost over 10lbs, I’m stressed all the time, and I can’t even spend time with my family because this is all I can think about…. 💔 I know many creators are afraid to speak publicly about situations like this, but I genuinely don’t know what else to do anymore. So, please. I’m asking for help. If the post came across your feed, please consider commenting, sharing, or tagging @TeamYouTube. Even a repost may help this situation reach someone willing to take a closer look. 🙏 I myself will continue doing my part for the community and help repost those on X that have been affected by this unfair demonetization. Thank you to everyone who is willing to support or has supported Mythical Arcana over the years. It truly means more than I can put into words. @TeamYouTube @YouTube @nealmohan @YouTubeCreators @UpdatesFromYT @reneritchie On the flip side. I want to say congratulations to all those who did manage to get their monetization back. I followed a lot of your journeys back, and it makes me hopeful to see creators finding success and justice. @MindlessPixels4 @AnimeVersus_X @stevenyjsong @MovieTalkNOW1 @AydinPaladin @HakariAltAcc @EternalMystYT @mortifyanims @punit_verma67 @Meowbananacat @Aiden213213 @DeliresdeMax @Earthfanboy0 @fantasy_images @TheBioloco @real_Monumental @ZzhuRafa @DanAnimation1 @SLHanimation @LoxtopiaAlicia Also, to the community of those who have been supporting creators and helping us, I give you a huge big thank you @FixYPP @ThatUmbrella @bear_ing @Kneon @thevivafrei @verbalriotshow @Nina7Infinity @MovieTalkNOW1 @CricketNewsHD4 @soulmatereading @XStrangeHistory @natecurtiss_yt @urban_studio_yt @1Skilaw @NoBotForSupport And @RayDocsProd #inauthenticcontent #youtube #youtubedemonetization #creatorsupport #youtubecreators #contentcreator #monetization #youtubepartner #fairreview #fixyoutube #youtubepolicy #supportcreators

English
1
0
0
12
Aaron Levie
Aaron Levie@levie·
CEOs are uniquely prone to AI psychosis because they’re sufficiently distant from the last mile of work that still has to happen to generate most value with AI. So when they play with AI, they see the happy path results, often not considering the next 10 or 20 things that have to happen to get sustainable results from agents. “Look I made this awesome product prototype”. Yes but you didn’t have to review the code before it went into production and fix a bunch of issues. “Look I generated a contract”. Yes but you didn’t verify all the terms before it goes out to the counterparty and didn’t have to wire up all the past contracts to work with. The best thing you can do as a CEO is to use AI a *ton* to figure out the real implications of agents in the enterprise, and come out the other side with an appreciation for both the upside and the real work that goes into them.
Michal Malewicz@michalmalewicz

CEOs are the most delusional about AI. Detached from reality.

English
268
649
6.2K
982.4K
avisre
avisre@avisre·
@thegoddessstyle That generational skepticism is important. Responsible AI has to include energy, provenance, consent, and whether the use case is valuable enough to justify the cost.
English
0
0
0
2
The Goddess Style
The Goddess Style@thegoddessstyle·
My 20 year old daughter who is planning to switch from Chemistry to Arts and Animation absolutely hates AI and the environmental impact on Earth. She gets upset when I use ChatGpt occasionally. I see their generation being more responsible than we give them credit for.
𐌁𐌉Ᏽ 𐌕𐌉𐌌𐌉@OrevaZSN

Teenagers have started calling AI art "boomer art" and consider it cringe, and YouTubers have stopped using AI-generated thumbnails because teenagers find them cringe and won't click on them. I honestly couldn't be happier.

English
1
0
1
27
avisre
avisre@avisre·
@moonlike_tech @anupamrjp Workflow is the key. AI coding tools get much more valuable when they understand the environment, hardware constraints, tests, and path from idea to working prototype.
English
0
0
0
8
moonlike
moonlike@moonlike_tech·
@anupamrjp Exort helps hardware builders go from idea to working prototype — not just code snippets. The future of AI coding tools is not only chat. It’s workflow. v0.2.0 is live: github.com/Razz19/Exort
English
1
0
0
12
🃏
🃏@anupamrjp·
Currently looking for founders building cool SaaS products 👇 Drop your startup + link.
English
11
0
9
405
avisre
avisre@avisre·
@ritsrnjn That is the real skill shift. AI coding moves effort from syntax to intent, constraints, review, tests, and deciding what good actually means.
English
0
0
0
6
rits
rits@ritsrnjn·
the thing nobody talks about with ai coding tools. you stop thinking in syntax. you start thinking in intent. that's not a shortcut. that's a different skill entirely.
English
1
0
0
14
avisre
avisre@avisre·
@therobertta_ Useful point. AI UX often works best when the interface invites exploration while still giving enough structure to prevent users from getting lost.
English
0
0
0
0