@MisbahSy

Experimenting in AI Founding DevRel @LiteLLM

Vancouver, Canada
Joined August 2015
Google's new tool turns your OpenClaw / Claude Code / Codex into a full-time assistant. And saves you on a bunch of SaaS subscriptions. nitter.cf/MisbahSy/status/202962…
21
75
854
217,036
Misbah Syed retweeted
wow!
Jev is definitely insane! What if you could just point and build a website? Uses json-render @ctatedev @LiteLLM gateway to call Jev
1
3
432
Here's how the full breakdown on my local machine
2
253
Jev is a super fast video editor! Edited this entire video in less than a minute, and here's how: First, I transcribed raw vid locally with word timestamps, cut it into 16 beats, then asked Jev these seven questions about every beat: does this line need a visual at all? which card template? (18 to pick from, premade) which text effect? which transition into it? how hard do we punch in on the face? which sound effect? which word gets emphasised? Plus five about the video as a whole. Style, colour, caption treatment, energy, progress bar. 117 questions answered by jev < 1 s ~ $0.0017 Using @typesafeai via @LiteLLM
6
14
2
56
4,510
Jev is definitely insane! What if you could just point and build a website? Uses json-render @ctatedev @LiteLLM gateway to call Jev
5
3
3
29
4,600
Measured on 19 documents / 155 pages, mistral-ocr-latest through a @LiteLLM gateway: OCR-only vs Jev-router 155 pages billed down to → 87 19 API requests down to → 13 35.6s → 20.7s (1.72x speed) $0.3100 → $0.1783 (1.74x savings) Repo: github.com/misbahsy/doc-rout…
4
1
24
1,723
Doc-OCR router using Jev @typesafeai A Jev-powered router that looks at a PDF page by page, decides which pages actually need OCR, extracts the rest locally. Result: save cost on # OCR pages + speed
11
34
3
483
26,208
TIL - Noul If you are gonna be using Jev, you might hear this too.
After co-inventing ChatGPT, I kept asking myself: why have superhuman chat models not led to AGI? I’ve spent the last 2 years in stealth building a new way to train models (RLCD), and a new type of frontier AI model that we are releasing today: Jev • 20-200x faster • 40-400x cheaper (w/ output tokens free) • Frontier composable intelligence optimized for decisions AFAICT the shortest path to AI-based economic revolution
4
529
41 common ai slop patterns from ai writing remove it with this one skill. @LiteLLM for multiple reviewers
anti-ai-slop: a 4 staged skill to remove any AI slop from your codex/cc writing. catches 41 different slop patterns: • ~120 ai words, ~100 phrases • em dashes, emoji bullets, title case headings • watermarks, negative parallelism, colon reveals, throat-clearing openers 4 gates: 1. sanitize invisible chars 2. remove common ai words/chars 3. an agent reviews original vs rewrite, loops back for failed reviews 4. optional: a parallel agent uses a different model to generate additional review + score Works with any of your AI agents.
1
6
10
1,281
A gift for your agents... (in case if you missed) github.com/misbahsy/anti-ai-…
anti-ai-slop: a 4 staged skill to remove any AI slop from your codex/cc writing. catches 41 different slop patterns: • ~120 ai words, ~100 phrases • em dashes, emoji bullets, title case headings • watermarks, negative parallelism, colon reveals, throat-clearing openers 4 gates: 1. sanitize invisible chars 2. remove common ai words/chars 3. an agent reviews original vs rewrite, loops back for failed reviews 4. optional: a parallel agent uses a different model to generate additional review + score Works with any of your AI agents.
4
9
996
anti-ai-slop: a 4 staged skill to remove any AI slop from your codex/cc writing. catches 41 different slop patterns: • ~120 ai words, ~100 phrases • em dashes, emoji bullets, title case headings • watermarks, negative parallelism, colon reveals, throat-clearing openers 4 gates: 1. sanitize invisible chars 2. remove common ai words/chars 3. an agent reviews original vs rewrite, loops back for failed reviews 4. optional: a parallel agent uses a different model to generate additional review + score Works with any of your AI agents.
6
5
2
28
4,658
Phew… Finally Gemini and XAI can catch up now
I agree with Dario that we need to pace the frontier. This has been a primary topic of discussions we've had at OpenAI in recent weeks. Committing to having independent evaluators with employee-like access is a great idea, and we will do the same. We'll have more to share soon.
2
7
423
For the sake of flykind, I trained a fly brain to find its way out of a room through open doors and windows. It scans the room, detects available exits, navigates around obstacles, and escapes. No more buzzing into walls and closed windows. Now I just need to pass the training to the rest of the fly community.
1
1
9
1,125
A simple gateway connected with AI tools can answer questions like: What models are we using most? How much do they cost? How many tokens used? Made this video to help connect @cursor_ai by @SpaceXAI with @LiteLLM and see the details of usage on a single dashboard. Link to the video: youtu.be/oNkMgelVsEA
3
1
8
1,460
I remember when Apple product launches were the most exciting launches in tech
iPhone Duo iPhone 18 Pro and iPhone 18 Pro Max Apple Watch Series 12 and Apple Watch Ultra 4 AirPods 5 … and Joz
3
525
Turn Astra into Sherlock and let it find what's missing on your business's website. Give this prompt to Astra, or drop your website URL. I'll run a few and DM the findings. PROMPT: Website: [YOUR WEBSITE URL] Be my startup’s growth and conversion auditor. Start with this website. Work out what we sell, who our customers are, and how someone moves from discovering us to getting value and becoming a paying customer. Your goal is to find the most useful improvements to our acquisition and conversion journey, backed by evidence from the actual website. Do not start with a long questionnaire. Investigate first, label your assumptions, and ask only if missing information materially changes the work. UNDERSTAND THE BUSINESS Identify: - The product or service and its business model. - The intended customers and their likely needs. - The main value proposition and supporting proof. - The primary conversion: purchase, signup, trial, demo, consultation, application, or another action. - Different customer journeys where relevant. Adapt the audit to this business. Do not assume every startup is SaaS or that everyone should book a sales call. FOLLOW THE CUSTOMER JOURNEY Use available browser, search, and page-inspection tools to explore the live website. Inspect the homepage, navigation, product/service pages, pricing, relevant landing pages, representative content, and conversion destinations. Follow important CTAs. Check where they actually lead and what the visitor encounters next, including third-party handoffs. Evaluate: - Can a new visitor quickly understand the offer and who it is for? - Is there a compelling reason to take the next step? - Are pricing, commitments, and expectations clear? - Is the proof relevant and credible? - Do CTA labels match their destinations? - Are forms and flows understandable and proportionate? - Are there broken links, confusing choices, or dead ends? - Is there a useful path for interested visitors who are not ready to buy? - Do important journeys work on mobile? Check critical links from a fresh visit, not only from within the destination page. If the site is large, prioritize commercially important paths and representative page types. State what you inspected and what you did not. SEPARATE FACTS FROM ASSUMPTIONS Classify findings as: 1. Confirmed defects, with reproducible evidence. 2. Friction hypotheses that need analytics or user testing. 3. Unknowns that require internal access. Record exact page URLs, CTA labels, destinations, and relevant observations. Identify what already works and should be preserved. Do not assume a third-party destination is a broken funnel. Do not claim that analytics, follow-up, or CRM integrations are missing just because they are not publicly visible. AUDIT SEARCH AND AI DISCOVERY Check representative pages for: - Crawlability and indexability signals. - Titles, descriptions, headings, and internal links. - Canonical URLs, robots directives, and sitemaps. - Clear, useful content that matches customer questions. - Accurate structured data aligned with visible content. - Consistent product claims, authorship, freshness, and credible evidence. - Opportunities to answer important buying and comparison questions. Use current authoritative guidance where needed. Do not invent rankings, traffic estimates, indexing outcomes, penalties, or AI-citation performance. Do not promise that a particular file, schema, or tactic guarantees visibility. CONNECT CONTENT TO CONVERSION Recommend where different acquisition messages should send people. For each important customer intent, specify: - The appropriate landing page or experience. - The primary next action. - Any useful secondary action for visitors not ready to convert. - An optional lead-capture offer, where appropriate. - The follow-up and outcome to measure. Do not send everyone to the homepage or recommend gating everything. Match the destination to the promise and the customer’s readiness. DESIGN A PRACTICAL MEASUREMENT PLAN Distinguish traffic, intent, activation, identified leads, qualified opportunities, purchases, and retention as relevant to the business. Propose: - Essential events and conversion milestones. - Attribution/context fields worth preserving. - Reliable confirmation of form submissions, bookings, or purchases. - Follow-up ownership. - Useful funnel and cohort metrics with clear denominators. - How to show missing data and unknown attribution. Do not equate button clicks with customers or form submissions with qualified demand. Respect consent and privacy. Do not recommend hidden tracking or unnecessary collection of personal data. PRIORITIZE THE IMPROVEMENTS For each recommendation, provide: - The evidence and affected customer journey. - Why it matters. - The proposed change. - Priority and rough effort. - Suggested owner and dependencies. - Acceptance criteria. - A success metric and relevant guardrails. Separate obvious repairs from experiments. Give me: - The five most important actions. - A first-week plan. - A practical 30/60/90-day roadmap. Do not invent revenue forecasts or guaranteed conversion uplift. CREATE A SHAREABLE REPORT Create a self-contained HTML report in the workspace, with supporting Markdown notes. Include: - Executive summary. - Business and audience understanding. - What already works. - Current customer journeys and CTA inventory. - Evidence-backed findings. - Search and AI-discovery opportunities. - Content-to-conversion recommendations. - Measurement plan. - Prioritized roadmap. - Sources, audit date, scope, and limitations. Include a transparent 0–100 public-readiness score with category weights and explanations. Adapt the categories to the business. Make clear that this is a heuristic assessment, not a measured conversion rate. Leave inaccessible dimensions unscored rather than treating unknowns as failures. Verify the report’s layout and links before delivering it. BOUNDARIES Audit only. Do not change the live website, submit forms, create accounts, book appointments, make purchases, contact anyone, or modify external systems without explicit permission. You may create local report files and use available subagents for independent research. Verify and consolidate their findings. If a tool or page is unavailable, explain the limitation and continue with accessible evidence. Finish with the key findings and the smallest set of additional data or read-only access needed to validate their business impact.
5
9
1,099
Misbah Syed retweeted
Thrilled to share @LiteLLM AutoRouter achieved 46% LOWER COST than Opus-5 on SWE-Bench w/ the SAME NUMBER SOLVED. Via Picking the model based on what subtask the agent is doing rather than difficulty. Read more here: docs.litellm.ai/blog/subtask…
4
4
10
615
Astra is insanely good at bringing vague ideas to life. Asked it to make a 3D version of a train flying through the Astra galaxy spiral with cosmic bodies in the background. Used a trick to burn fewer tokens; simply asked Astra to check its work frequently. It repeatedly checked intermediate renders to ensure it stayed on track. The model is surprisingly fast. This was on medium reasoning, where it used @threejs and @HyperFrames_ to generate the video.
5
2
13
953
PS: This was just a random idea as how to showcase the launch of Astra and its support on @LiteLLM
1
2
136