Day 76/100 of #100DaysWithAI 📦
A sandbox isn't a cut leg. Claude's code interpreter allowlists github.com, and Willison sees exfiltration routes in that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
A sandbox isn't a cut leg. Claude's code interpreter allowlists github.com, and Willison sees exfiltration routes in that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
September 25, 2026 at 3:51 AM
Day 76/100 of #100DaysWithAI 📦
A sandbox isn't a cut leg. Claude's code interpreter allowlists github.com, and Willison sees exfiltration routes in that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
A sandbox isn't a cut leg. Claude's code interpreter allowlists github.com, and Willison sees exfiltration routes in that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
Day 75/100 of #100DaysWithAI 🔱
75 days in, the rule I keep: private data + untrusted content + a way out = an agent that can be talked into leaking. Cut one leg.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
75 days in, the rule I keep: private data + untrusted content + a way out = an agent that can be talked into leaking. Cut one leg.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
September 24, 2026 at 2:59 AM
Day 75/100 of #100DaysWithAI 🔱
75 days in, the rule I keep: private data + untrusted content + a way out = an agent that can be talked into leaking. Cut one leg.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
75 days in, the rule I keep: private data + untrusted content + a way out = an agent that can be talked into leaking. Cut one leg.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISecurity
🤝 Claude Code-assisted, reviewed by me.
Day 74/100 of #100DaysWithAI 🔁
14 LLMs, 5 runs each: 90-98% perfect self-agreement. On actual market moves, all at chance. Consistent is not the same as correct.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
14 LLMs, 5 runs each: 90-98% perfect self-agreement. On actual market moves, all at chance. Consistent is not the same as correct.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
September 23, 2026 at 2:54 AM
Day 74/100 of #100DaysWithAI 🔁
14 LLMs, 5 runs each: 90-98% perfect self-agreement. On actual market moves, all at chance. Consistent is not the same as correct.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
14 LLMs, 5 runs each: 90-98% perfect self-agreement. On actual market moves, all at chance. Consistent is not the same as correct.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
Day 69/100 of #100DaysWithAI 📈
RL lifts the problems a model already solves and barely moves the hard ones - the rich get richer. NGU samples until one lands.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #MachineLearning
🤝 Claude Code-assisted, reviewed by me.
RL lifts the problems a model already solves and barely moves the hard ones - the rich get richer. NGU samples until one lands.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #MachineLearning
🤝 Claude Code-assisted, reviewed by me.
September 18, 2026 at 4:10 AM
Day 69/100 of #100DaysWithAI 📈
RL lifts the problems a model already solves and barely moves the hard ones - the rich get richer. NGU samples until one lands.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #MachineLearning
🤝 Claude Code-assisted, reviewed by me.
RL lifts the problems a model already solves and barely moves the hard ones - the rich get richer. NGU samples until one lands.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #MachineLearning
🤝 Claude Code-assisted, reviewed by me.
Day 73/100 of #100DaysWithAI 🎲
An LLM judge that prints "4" discards how sure it was. G-Eval weights it by token log-probs - the doubt becomes part of the number.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
An LLM judge that prints "4" discards how sure it was. G-Eval weights it by token log-probs - the doubt becomes part of the number.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
September 22, 2026 at 2:50 AM
Day 73/100 of #100DaysWithAI 🎲
An LLM judge that prints "4" discards how sure it was. G-Eval weights it by token log-probs - the doubt becomes part of the number.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
An LLM judge that prints "4" discards how sure it was. G-Eval weights it by token log-probs - the doubt becomes part of the number.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
Day 14/100 of #100DaysWithAI 🧩
Two days I built the agent loop by hand - today, when not to. Most "agents" are better as workflows: chaining, routing, parallelisation. Reach for autonomy only when the steps are unpredictable.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Two days I built the agent loop by hand - today, when not to. Most "agents" are better as workflows: chaining, routing, parallelisation. Reach for autonomy only when the steps are unpredictable.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
July 25, 2026 at 4:06 AM
Day 14/100 of #100DaysWithAI 🧩
Two days I built the agent loop by hand - today, when not to. Most "agents" are better as workflows: chaining, routing, parallelisation. Reach for autonomy only when the steps are unpredictable.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Two days I built the agent loop by hand - today, when not to. Most "agents" are better as workflows: chaining, routing, parallelisation. Reach for autonomy only when the steps are unpredictable.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Day 72/100 of #100DaysWithAI 📏
Rating LLM output 1-5 feels rigorous. Annotators drift to the middle. Binary pass/fail forces the call - and names the failure.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
Rating LLM output 1-5 feels rigorous. Annotators drift to the middle. Binary pass/fail forces the call - and names the failure.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
September 21, 2026 at 3:03 AM
Day 72/100 of #100DaysWithAI 📏
Rating LLM output 1-5 feels rigorous. Annotators drift to the middle. Binary pass/fail forces the call - and names the failure.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
Rating LLM output 1-5 feels rigorous. Annotators drift to the middle. Binary pass/fail forces the call - and names the failure.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AIEvals
🤝 Claude Code-assisted, reviewed by me.
Day 68/100 of #100DaysWithAI 🧰
Someone forked Claude Code to run on any model. What survives the swap - tools, agents, MCP, slash commands - is the actual product.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Someone forked Claude Code to run on any model. What survives the swap - tools, agents, MCP, slash commands - is the actual product.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
September 17, 2026 at 3:52 AM
Day 68/100 of #100DaysWithAI 🧰
Someone forked Claude Code to run on any model. What survives the swap - tools, agents, MCP, slash commands - is the actual product.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Someone forked Claude Code to run on any model. What survives the swap - tools, agents, MCP, slash commands - is the actual product.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Day 67/100 of #100DaysWithAI 🧬
DeepMind ran AlphaGenome over every possible DNA letter change and shipped the answers. 9B variants, 1PB. A model served as a table.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #DeepLearning
🤝 Claude Code-assisted, reviewed by me.
DeepMind ran AlphaGenome over every possible DNA letter change and shipped the answers. 9B variants, 1PB. A model served as a table.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #DeepLearning
🤝 Claude Code-assisted, reviewed by me.
September 16, 2026 at 3:56 AM
Day 67/100 of #100DaysWithAI 🧬
DeepMind ran AlphaGenome over every possible DNA letter change and shipped the answers. 9B variants, 1PB. A model served as a table.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #DeepLearning
🤝 Claude Code-assisted, reviewed by me.
DeepMind ran AlphaGenome over every possible DNA letter change and shipped the answers. 9B variants, 1PB. A model served as a table.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #DeepLearning
🤝 Claude Code-assisted, reviewed by me.
Day 71/100 of #100DaysWithAI 🔁
Cursor's agent docs have no error-handling section. Just a terminal that monitors output. Your error messages are the feedback loop.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
Cursor's agent docs have no error-handling section. Just a terminal that monitors output. Your error messages are the feedback loop.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
September 20, 2026 at 3:57 AM
Day 71/100 of #100DaysWithAI 🔁
Cursor's agent docs have no error-handling section. Just a terminal that monitors output. Your error messages are the feedback loop.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
Cursor's agent docs have no error-handling section. Just a terminal that monitors output. Your error messages are the feedback loop.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
Day 42/100 of #100DaysWithAI ✋
Cline's default is a stop. Edits and commands wait for you; Plan mode cannot write until you switch. Open source, any model, your key.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #OpenSource
🤝 Claude Code-assisted, reviewed by me.
Cline's default is a stop. Edits and commands wait for you; Plan mode cannot write until you switch. Open source, any model, your key.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #OpenSource
🤝 Claude Code-assisted, reviewed by me.
August 22, 2026 at 4:22 AM
Day 42/100 of #100DaysWithAI ✋
Cline's default is a stop. Edits and commands wait for you; Plan mode cannot write until you switch. Open source, any model, your key.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #OpenSource
🤝 Claude Code-assisted, reviewed by me.
Cline's default is a stop. Edits and commands wait for you; Plan mode cannot write until you switch. Open source, any model, your key.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #OpenSource
🤝 Claude Code-assisted, reviewed by me.
Day 38/100 of #100DaysWithAI 📚
Four Claude products say "build an agent". Only the SDK runs Claude Code's own loop in your process - loading the .claude/ skills, hooks and permissions you already wrote.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #ClaudeCode
Four Claude products say "build an agent". Only the SDK runs Claude Code's own loop in your process - loading the .claude/ skills, hooks and permissions you already wrote.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #ClaudeCode
August 18, 2026 at 4:25 AM
Day 38/100 of #100DaysWithAI 📚
Four Claude products say "build an agent". Only the SDK runs Claude Code's own loop in your process - loading the .claude/ skills, hooks and permissions you already wrote.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #ClaudeCode
Four Claude products say "build an agent". Only the SDK runs Claude Code's own loop in your process - loading the .claude/ skills, hooks and permissions you already wrote.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #ClaudeCode
Day 52/100 of #100DaysWithAI 🔍
The reviewer is a separate agent: it gets the draft, not the reasons for it. Self-review shares the reasons, so it spares them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
The reviewer is a separate agent: it gets the draft, not the reasons for it. Self-review shares the reasons, so it spares them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
September 1, 2026 at 4:49 AM
Day 52/100 of #100DaysWithAI 🔍
The reviewer is a separate agent: it gets the draft, not the reasons for it. Self-review shares the reasons, so it spares them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
The reviewer is a separate agent: it gets the draft, not the reasons for it. Self-review shares the reasons, so it spares them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Day 65/100 of #100DaysWithAI 🔍
100 agents. One found a grader exploit, 24% raised the alarm and still lost - no way to delete a fake proof. Design, not morals.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
100 agents. One found a grader exploit, 24% raised the alarm and still lost - no way to delete a fake proof. Design, not morals.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
September 14, 2026 at 4:44 AM
Day 65/100 of #100DaysWithAI 🔍
100 agents. One found a grader exploit, 24% raised the alarm and still lost - no way to delete a fake proof. Design, not morals.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
100 agents. One found a grader exploit, 24% raised the alarm and still lost - no way to delete a fake proof. Design, not morals.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Day 70/100 of #100DaysWithAI 🔐
Self-hosting your AI chat moves the privacy line, it does not erase it. The prompts still leave for whichever provider you picked.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #GenAI
🤝 Claude Code-assisted, reviewed by me.
Self-hosting your AI chat moves the privacy line, it does not erase it. The prompts still leave for whichever provider you picked.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #GenAI
🤝 Claude Code-assisted, reviewed by me.
September 19, 2026 at 3:02 AM
Day 70/100 of #100DaysWithAI 🔐
Self-hosting your AI chat moves the privacy line, it does not erase it. The prompts still leave for whichever provider you picked.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #GenAI
🤝 Claude Code-assisted, reviewed by me.
Self-hosting your AI chat moves the privacy line, it does not erase it. The prompts still leave for whichever provider you picked.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #GenAI
🤝 Claude Code-assisted, reviewed by me.
Day 62/100 of #100DaysWithAI 📈
AI agents lifted coding activity 180%. Projects rose 50%, releases 30%, downloads didn't follow. The bottleneck moved.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #SoftwareEngineering
🤝 Claude Code-assisted, reviewed by me.
AI agents lifted coding activity 180%. Projects rose 50%, releases 30%, downloads didn't follow. The bottleneck moved.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #SoftwareEngineering
🤝 Claude Code-assisted, reviewed by me.
September 11, 2026 at 4:52 AM
Day 62/100 of #100DaysWithAI 📈
AI agents lifted coding activity 180%. Projects rose 50%, releases 30%, downloads didn't follow. The bottleneck moved.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #SoftwareEngineering
🤝 Claude Code-assisted, reviewed by me.
AI agents lifted coding activity 180%. Projects rose 50%, releases 30%, downloads didn't follow. The bottleneck moved.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #SoftwareEngineering
🤝 Claude Code-assisted, reviewed by me.
Day 58/100 of #100DaysWithAI 🚨
OpenAI's root cause names no jailbreak. Three properties compounding: powerful, persistent, collaborative. Only one needs two agents.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISafety
🤝 Claude Code-assisted, reviewed by me.
OpenAI's root cause names no jailbreak. Three properties compounding: powerful, persistent, collaborative. Only one needs two agents.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISafety
🤝 Claude Code-assisted, reviewed by me.
September 7, 2026 at 3:39 AM
Day 58/100 of #100DaysWithAI 🚨
OpenAI's root cause names no jailbreak. Three properties compounding: powerful, persistent, collaborative. Only one needs two agents.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISafety
🤝 Claude Code-assisted, reviewed by me.
OpenAI's root cause names no jailbreak. Three properties compounding: powerful, persistent, collaborative. Only one needs two agents.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AISafety
🤝 Claude Code-assisted, reviewed by me.
Day 15/100 of #100DaysWithAI 🧱
A third from-scratch agent, and the honest one: it adds memory across turns, then names what's still missing - streaming, retries, structured output. The scaffold is easy; production is the roadmap.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
A third from-scratch agent, and the honest one: it adds memory across turns, then names what's still missing - streaming, retries, structured output. The scaffold is easy; production is the roadmap.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
July 26, 2026 at 4:06 AM
Day 15/100 of #100DaysWithAI 🧱
A third from-scratch agent, and the honest one: it adds memory across turns, then names what's still missing - streaming, retries, structured output. The scaffold is easy; production is the roadmap.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
A third from-scratch agent, and the honest one: it adds memory across turns, then names what's still missing - streaming, retries, structured output. The scaffold is easy; production is the roadmap.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
Day 5/100 of #100DaysWithAI 🏗️
OpenAI's field note: teams run 15+ distinct tools fine yet struggle with 10 overlapping ones - overlap, not count, kills agents. Max out one agent before going multi-agent.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
OpenAI's field note: teams run 15+ distinct tools fine yet struggle with 10 overlapping ones - overlap, not count, kills agents. Max out one agent before going multi-agent.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
July 16, 2026 at 6:00 AM
Day 5/100 of #100DaysWithAI 🏗️
OpenAI's field note: teams run 15+ distinct tools fine yet struggle with 10 overlapping ones - overlap, not count, kills agents. Max out one agent before going multi-agent.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
OpenAI's field note: teams run 15+ distinct tools fine yet struggle with 10 overlapping ones - overlap, not count, kills agents. Max out one agent before going multi-agent.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Day 7/100 of #100DaysWithAI 🧭
Anthropic's Fable 5 prompting guide, inverted: skills written for older models are often too prescriptive and can DEGRADE output. Migration isn't adding instructions - it's deleting them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
Anthropic's Fable 5 prompting guide, inverted: skills written for older models are often too prescriptive and can DEGRADE output. Migration isn't adding instructions - it's deleting them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
July 18, 2026 at 5:01 AM
Day 7/100 of #100DaysWithAI 🧭
Anthropic's Fable 5 prompting guide, inverted: skills written for older models are often too prescriptive and can DEGRADE output. Migration isn't adding instructions - it's deleting them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
Anthropic's Fable 5 prompting guide, inverted: skills written for older models are often too prescriptive and can DEGRADE output. Migration isn't adding instructions - it's deleting them.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #LLM
Day 45/100 of #100DaysWithAI 🧰
Matt Pocock's skills split on one axis: who may invoke them. Orchestrators may call disciplines, never each other.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Matt Pocock's skills split on one axis: who may invoke them. Orchestrators may call disciplines, never each other.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
August 25, 2026 at 3:44 AM
Day 45/100 of #100DaysWithAI 🧰
Matt Pocock's skills split on one axis: who may invoke them. Orchestrators may call disciplines, never each other.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Matt Pocock's skills split on one axis: who may invoke them. Orchestrators may call disciplines, never each other.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #AgenticAI
🤝 Claude Code-assisted, reviewed by me.
Day 24/100 of #100DaysWithAI 🐫
CaMeL's trick: the LLM that reads your untrusted data has no tools, and the one holding the tools never sees that data. Security bought with architecture, not with a better filter.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
CaMeL's trick: the LLM that reads your untrusted data has no tools, and the one holding the tools never sees that data. Security bought with architecture, not with a better filter.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
August 4, 2026 at 5:13 AM
Day 24/100 of #100DaysWithAI 🐫
CaMeL's trick: the LLM that reads your untrusted data has no tools, and the one holding the tools never sees that data. Security bought with architecture, not with a better filter.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
CaMeL's trick: the LLM that reads your untrusted data has no tools, and the one holding the tools never sees that data. Security bought with architecture, not with a better filter.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Day 11/100 of #100DaysWithAI 🔌
Wiring every agent to every tool by hand is an N×M problem. MCP makes it N+M - one protocol the model discovers at runtime. Adoption curves follow arithmetic like that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #MCP
Wiring every agent to every tool by hand is an N×M problem. MCP makes it N+M - one protocol the model discovers at runtime. Adoption curves follow arithmetic like that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #MCP
July 22, 2026 at 4:52 AM
Day 11/100 of #100DaysWithAI 🔌
Wiring every agent to every tool by hand is an N×M problem. MCP makes it N+M - one protocol the model discovers at runtime. Adoption curves follow arithmetic like that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #MCP
Wiring every agent to every tool by hand is an N×M problem. MCP makes it N+M - one protocol the model discovers at runtime. Adoption curves follow arithmetic like that.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #MCP
Day 21/100 of #100DaysWithAI 🎛️
Stop forking an agent per customer. Make its config a function: instructions, model, tools and memory each read a typed request context, so one agent resolves itself per call.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Stop forking an agent per customer. Make its config a function: instructions, model, tools and memory each read a typed request context, so one agent resolves itself per call.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
August 1, 2026 at 5:00 AM
Day 21/100 of #100DaysWithAI 🎛️
Stop forking an agent per customer. Make its config a function: instructions, model, tools and memory each read a typed request context, so one agent resolves itself per call.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Stop forking an agent per customer. Make its config a function: instructions, model, tools and memory each read a typed request context, so one agent resolves itself per call.
🔗 https://github.com/aurimas13/100-Days-With-AI
#AI #AgenticAI
Day 43/100 of #100DaysWithAI ⌨️
Three coding agents, three days, all with one primitive: one mode that may write, one that may not. opencode: terminal, MIT licensed.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
Three coding agents, three days, all with one primitive: one mode that may write, one that may not. opencode: terminal, MIT licensed.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
August 23, 2026 at 4:35 AM
Day 43/100 of #100DaysWithAI ⌨️
Three coding agents, three days, all with one primitive: one mode that may write, one that may not. opencode: terminal, MIT licensed.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.
Three coding agents, three days, all with one primitive: one mode that may write, one that may not. opencode: terminal, MIT licensed.
🔗 https://github.com/aurimas13/100-Days-With-AI
#LearningInPublic #CodingAgents
🤝 Claude Code-assisted, reviewed by me.